research paper crime and media

The Relationship between Media Consumption and Fear of Crime: A Comprehensive Review

JIB Journal., 11(1), 56–62, 2023. doi: 10.34050/jib.v11i1.26280

7 Pages Posted: 11 May 2023 Last revised: 29 Aug 2023

Banuka De Silva

General Sir John Kotelawala Defence University

Date Written: April 23, 2023

This review comprehensively analyses the relationship between media consumption and fear of crime. Most reviewed studies indicate a positive correlation between media consumption and fear of crime, which can be attributed to the media's tendency to focus on sensationalised and dramatic crimes and the media's effect on individuals' perceptions of reality; however, several studies have found a negative correlation, explained by desensitisation, and heightened awareness among media consumers. Media literacy programmes and responsible reporting are critical to mitigate the public's fear of crime. By promoting media literacy, individuals can better understand the prevalence of crime and recognise the sensationalised nature of media coverage of crime. Media organisations must report crime-related news accurately and objectively, without exaggerating events, to promote a less distorted view of crime in society. The systematic approach employed in this review and the specific search strategy provides a comprehensive and trustworthy overview of the existing research on the relationship between media consumption and fear of crime, despite the limitations of the search strategy. Additional research is needed to understand better the underlying mechanisms and contextual factors that can influence the relationship; nonetheless, this analysis offers valuable insights into the effect of media consumption on public perceptions of crime and the resulting implications for public safety. Media organisations and individuals must take a responsible approach to consume and reporting crime-related news to promote a more accurate and less distorted view of crime in society.

Keywords: media consumption, fear of crime, media literacy, responsible reporting, sensationalism,

Suggested Citation: Suggested Citation

K.B.N. De Silva (Contact Author)

General sir john kotelawala defence university ( email ).

Kandawala Rd, Ratmalana, 10390 Sri Lanka

Do you have a job opening that you would like to promote on SSRN?

Paper statistics, related ejournals, criminology ejournal.

Subscribe to this fee journal for more curated articles on this topic

Law & Society: Public Law - Crime, Criminal Law, & Punishment eJournal

Social & personality psychology ejournal, mass communication & popular culture ejournal, psychology research methods ejournal, sociology of media & technology ejournal.

Tools and Resources
Customer Services
Corrections
Crime, Media, and Popular Culture
Criminal Behavior
Criminological Theory
Critical Criminology
Geography of Crime
International Crime
Juvenile Justice
Prevention/Public Policy
Race, Ethnicity, and Crime
Research Methods
Victimology/Criminal Victimization
White Collar Crime
Women, Crime, and Justice
Share This Facebook LinkedIn Twitter

Sign in to an additional subscriber account

Article contents, content analysis in the study of crime, media, and popular culture.

Lisa A. Kort-Butler Lisa A. Kort-Butler Department of Sociology, University of Nebraska–Lincoln
https://doi.org/10.1093/acrefore/9780190264079.013.23
Published online: 29 September 2016

Content analysis is considered both a quantitative and a qualitative research method. The overarching goal of much of the research using this method is to demonstrate and understand how crime, deviance, and social control are represented in the media and popular culture. Unlike surveys of public opinions about crime issues, which seek to know what people think or feel about crime, content analysis of media and popular culture aims to reveal a culture’s story about crime. Unlike research that examines how individuals’ patterns of media consumption shape their attitudes about crime and control, content analysis appraises the meaning and messages within the media sources themselves. Media and popular culture sources are viewed as repositories of cultural knowledge, which capture past and present ideas about crime, while creating and reinforcing a culture’s shared understanding about crime.

In content analysis, media and popular culture portrayals of crime issues are the primary sources of data. These portrayals include a range of sources, such as newspapers, movies, television programs, advertisements, comic books, novels, video games, and Internet content. Depending on their research questions, researchers draw samples from their selected sources, usually with additional selection boundaries, such as timeframe, genre, and topic (e.g., movies about gangs released from 1960 to 1990).

There are two primary approaches to conducting content analysis. In quantitative forms of content analysis, researchers code and count the occurrence of elements designated by the researcher prior to the study (e.g., the number of times a violent act occurs). In qualitative forms of content analysis, the researchers focus on the narrative, using an open-ended protocol to record information. The approaches are complementary, as each reveals unique yet overlapping concepts crucial to understanding how the media and popular culture produce and reproduce ideas about crime.

content analysis
qualitative methods

Defining Content Analysis

As a research method, content analysis exists somewhere between purely quantitative and purely qualitative. In the study of crime in the media, research ranges from studies that count or otherwise quantify texts for the purpose of statistical analysis to studies that explore presentation and representation of crime-related issues. Even in those quantitatively oriented studies, results are given qualitative consideration. Increasingly, in the criminological study of media and popular culture, content analysis is typically viewed as a qualitative methodology.

Content analysis is more than watching TV or movies, or reading newspapers or comics, and then reporting what is presented in the medium. How the story is told and how characters are portrayed are often more telling than are specific plot points. Content analysis requires systematically watching or reading with an analytical and critical eye, going beyond what is presented and looking for deeper meanings and messages to which media consumers are exposed. This exposure contributes to the social construction of crime and deviance, that is, to people’s beliefs about what is deviant, who is criminal, and how to control crime. The media captures and frames the broader cultural story about criminal justice. The primary purpose of content analysis in the study of crime and justice has evolved from identifying the prevalence of the topic or terms under study into revealing the cultural frames. The results from content analysis, then, offer evidence that allows for a more critical appraisal of how crime and justice are socially constructed.

The past 40 years have seen substantial growth in the application of content analysis to a range of issues, including crime and violence (Neuendorf, 2002 ), moving beyond text on the page to “text” in visual and moving images. Indeed, one of the earliest studies to employ the method, the Payne Fund Studies, coded for violence and other content in films in the 1930s. As the century progressed and attention shifted to violence in television, content analysis became a core methodology of the Cultural Indicators Project (Gerbner & Gross, 1976 ; Morgan & Shanahan, 2010 ). This project has influenced media research since the 1970s, including the National Television Violence Study of the later 1990s (Smith et al., 1998 ).

As research has progressed, however, scholars called for greater attention to the context in which the content is presented, arguing that an act or an incident could not be fully understood without referring to the circumstances of its presentation in media or the broader socio-cultural context. Such work, coupled with expanding opportunities for consumers to encounter crime-related content across a variety of media sources, also stimulated analyses that placed more emphasis on the latent content itself. That is, some research looks beyond the action to the less obvious, but still critical, message and meaning being produced and reproduced in the media and popular culture. The advent of academic journals such as Crime, Media, Culture , Journal of Criminal Justice and Popular Culture , and the recent Journal of Qualitative Criminal Justice and Criminology also speak to the emergence of content analysis and other qualitative techniques in the study of crime and social control.

Sampling the Media Universe

For those studying crime issues presented in media and popular culture, there is a wide array of text-based sources, including novels, textbooks, newspapers, magazines, and comic books and graphic novels. There is also a wide array of audio-visual sources, including movies, television, and video games, each with a myriad of genres and formats. Music, in lyrics, video, and performance, is yet another source. Finally, the rise of electronic and social media further broadens the range of sources, from traditional news sources to Twitter conversations to YouTube content to user-generated forums like Reddit. The type and genre of media to be studied are often identified as research questions are developed.

As with most social research, it is often not feasible to examine an entire population of media texts or sources. For example, even if one could access every copy of comic books featuring Batman, it is likely impractical, due to constraints on time or resources, to read and code hundreds, if not thousands, of books. On the other hand, one could watch and code every cinematic release featuring Batman (Bosch, 2016 ). Thus, the decision between reviewing an entire population or a sample of the population is driven by both research questions and practical considerations.

With research questions and practical considerations in mind, sampling entails additional decision points. For example, suppose one is interested in news presentations of crime in editorials or commentaries. First, one needs to decide among newspapers, news magazines, televisions news, or Internet news. If U.S. newspapers are selected, then one needs to determine the sampling frame, including the time period of interest and which papers to select. Will papers be randomly selected from the universe of U.S. papers or from papers with a certain circulation level? Will there be a degree of stratification, such as random samples from designated geographic units? Or, will newspapers be selectively chosen based on other research interests, such as tracking a specific event in a specific location? If, however, one wants to compare across media types, then similar decisions need to be made that can be applied to each type. Within each source, there may be several stories, editorials, or commentaries, so researchers need to decide whether to review all of them, or more practically, decide how to sample among them, necessitating another round of sampling decisions.

In sum, as this brief example illustrates, sampling for the purpose of content analysis entails a good deal of complexity. Regardless of design, samples should be selected so that they reasonably represent the population and yield sufficient numbers for analysis. Researchers should also take care to record all decision points, so that the sample can be replicated by others.

Quantitative Content Analysis

Originally developed for use with written texts, quantitative content analysis (QCA) aims to distill the many words presented in a text into meaningful categories. These categories can then be treated as variables, allowing for a descriptive interpretation of the texts, or functioning as variables in statistical analyses. QCA has expanded beyond the written word to many other types of media, but the basic principal of classifying larger content into smaller categories remains at the core of the method. Through analysis of how these categories inter-relate with each other and intersect with the broader cultural context, the goal is to discover how materials communicate meaning and what meaning is communicated.

Basic Methodology of QCA

A coding schema is central to the method. As with sampling, the development of a coding schema is driven by precedents in the research or theoretical literature, and by the teams’ research questions. Weber ( 1990 ) outlines several basic steps in the coding process. The first step is to define recording units, that is, whether coders should attend to certain words, phrases, images, or overarching themes of a passage or piece. Depending on research questions, recording units often are some combination of these or other units. For example, in a television show, researchers may want to know what words were in the dialogue and also the overarching theme of the conversation between characters.

The second step in developing a coding schema is to define categories. One may think of these as boxes to be marked on a rubric, even if computer software is aiding in the coding process. The categories may be defined narrowly or broadly. For example, the presence of “violence” between two characters may be narrowly defined by physical contact or broadly defined to include abusive language. The categories may be mutually exclusive, or an incident may be coded into multiple categories. This decision may be influenced by the analytic intent; basic statistical assumptions are violated when categories are not exclusive.

Once a preliminary rubric is established, the third step is a pre-test, in which a subsample is coded. The pre-test process should reveal where categories may need further refinement or where coding rules require additional clarity. The pre-test also produces information regarding the accuracy or correctness with which human coders or computerized coding programs are classifying the text. If accuracy is low, the coding rules should be revised. Step four, then, is revising the coding schema. The fifth step is to pre-test again. This process should continue until the coding process, whether human- or computer-coded, yields an adequate level of accuracy.

The sixth step is to code the full sample using the established schema. Following data collection, the final step of the coding process is checking the achieved accuracy of the human coders or the computer program. Individual human coders may fatigue over time, thus making more mistakes, or their interpretation of categories may shift slightly over time, resulting in misclassification. Computerized output should be reviewed to confirm whether code rules were applied correctly. During the process, for example, the program may encounter text combinations not anticipated by the programmers or not present in the pre-test, resulting in misclassification.

As with other forms of measurement, issues of reliability and validity may emerge in QCA (Neuendorf, 2002 ). Particularly with several human coders, a primary issue is inter-rater reliability. In brief, inter-rater reliability is the extent to which different people code the same text in the same way. Differences, for example, may occur when coding rules or categories are not clear, or when there are cognitive differences across coders. The pre-test process and adequate coder training may reduce these differences, but inter-rater reliability should also be assessed at the end of coding. Various statistical tests exist to assess inter-rater reliability.

Validity can broadly be divided into internal and external. Internal validity refers to the match between concepts and their operationalized definitions in variables. There is no parallel statistical metric to assess internal validity, but there are several dimensions that researchers may consider. External validity, in contrast, refers to the generalizability and replicability of the results generated by a measure. Breadth and representativeness of the sample improve generalizability, while a full accounting of the procedures of the coding and variable creation improve replicability.

Analysis and Interpretation in QCA

Analyzing data generated by the coding process can take many forms. Again, the analysis of the data is driven by existing theory and the established research questions. Once data are collected, however, researchers fully quantify the data by creating variables from the coded data that are most meaningful for the hypotheses they want to explore and the analyses they want to conduct. These analyses may range from completely descriptive in nature to mean-difference or correlational tests to multivariate regression models.

In interpreting the analytical results, researchers bring the accumulated evidence to bear on the research questions, determining what story their results tell about the texts and their content. Regardless of the analytical technique used in QCA, any interpretation of quantified content must be corroborated by reference to the original texts. That is, researchers should compare their interpretation of the data to a subsample of their source documents. If the story of the data analysis reasonably represents the story within the documents, the interpretations of analytical results are not just products of classification schemes or statistical techniques. If the stories do not match, then reconsideration of the analysis or interpretation is necessary. In short, although QCA aims to quantify what could be considered qualitative information, it nonetheless retains a portion of qualitative art in the final interpretation of the data.

Example of QCA

One example of QCA is Britto and Dabney’s ( 2010 ) analysis of justice issues on political talk shows, which illustrates the coding and analysis process. Britto and Dabney were interested in crime content on these programs and, in particular, how the content was politicized. They selected the central primetime talk show across each of the three major cable news networks. Over the course of six months, they randomly selected one day per week to record the shows. They chose this approach for two reasons: to avoid one particular news story dominating conversation (e.g., a high-profile crime) and to ensure they were watching a “typical” example of the show. Coding was performed at two levels of analysis. At the program level, the schema included numeric coding categories for the number of segments in an episode, speaking time given to guests, and racial/ethnic and gender characteristics of offenders and victims in crime stories. At the individual level of analysis, the schema included codes for guest characteristics and guests’ interactions with hosts. These categories became variables in the analyses. Coders received four hours of training, which included a discussion of how concepts were operationalized, a practice coding session, and the follow-up discussion. During data collection, coders were instructed to watch their assigned episodes at least twice in order to code at each analytic level separately.

Britto and Dabney’s statistical analysis of their enumerated data began with a presentation of guest profiles, comparing population numbers to the demographic characteristics of guests, and then comparing general show guests to guests in justice-related segments. The analysis continued with a description of the amount of justice-related content on each show, then compared shows to each other. To address their major research questions, the analysis examined the interactions of hosts with their guests, using chi-square tests to reveal differences across guests’ political persuasions. The analysis also presented ratios of offender and victim characteristics, comparing these ratios across shows and to official United States crime data. Britto and Dabney used these statistical presentations as evidence in answering their research questions; however, they also corroborated their results by presenting qualitative descriptions of each of the source programs.

Other examples of QCA include Welch, Weber, and Edwards’ ( 2000 ) analysis of corrections debates in the New York Times ; Cavender, Bond-Maupin, and Jurik’s ( 1999) analysis of gender in reality crime television; and Chiricos and Eschholz’s ( 2002 ) analysis of crime and race/ethnicity in local television news.

Ethnographic Content Analysis

The quantitative approach to content analysis, while useful, may result in removing the coded content from the context surrounding it (Muzzatti, 2006 ). For example, a “violent action” in a television program may be counted, but the scene or setting in which it occurred, the offender’s motive for the violent act, the victim’s reaction, and other visual or auditory detail are lost. To address the shortcomings of QCA, Altheide ( 1987 ) proposed for ethnographic content analysis (ECA).

Like QCA, the goal of ECA is to discover how materials communicate meaning; however, in ECA, meaning is assumed to be present in various modalities, such as text, format, visual and auditory style, and in the positioning of one piece of information among others. Although this approach may involve some degree of enumeration, the emphasis is on descriptive and conceptual data, similar to what one might record in an observational study. The procedures for data collection, analysis, and interpretation are designed to be reflexive, with the researcher constantly engaged in the material, comparing observations as the process unfolds, and attentive to conceptual and theoretical nuances as they arise.

Basic Methodology in ECA

The basic methodology of ECA is outlined by Altheide and Schneider ( 2013 ). A review of the literature should guide the selection of the specific problem or issue to examine, as well as the medium to examine. Once researchers decide on a medium, they should learn about its production process and context. For example, if researchers are interested in studying comic books, they should become familiar with how comics are created, developed, and disseminated. Next, researchers should become familiar with several key examples to understand elements of formatting or general patterns in how the text and imagery are presented. Together, an understanding of the production process and the patterning within the source material provide essential background for creating a data collection protocol.

Constructing an ECA protocol is similar to developing a coding schema in QCA in the sense that the protocol is developed, tested, and re-tested. The format of an ECA protocol, however, is oriented conceptually rather than categorically. A protocol is designed as a means by which to query the information source—to ask questions, capture definitions, understand meanings, and reveal processes. Theoretically derived concepts, drawn from the research literature and from evaluation of key examples, guide the creation of a preliminary protocol. The protocol should have appropriate preset coding categories, (e.g., citation, date, length of text), but most of the conceptual categories should be open-ended for the researcher to input pertinent information, whether it be a count, a quotation, or a narrative description. During protocol development, new conceptual classifications are expected to emerge; as a result, original categories may be modified, and new categories may be incorporated into the protocol.

Data collection using the established protocol entails populating the categories with a wealth of descriptive examples. Throughout data collection, the researcher remains attuned to the process, such as including additional notes about emergent themes or observations about how pieces in the sample connect to each other. Indeed, Altheide and Schneider ( 2013 ) recommend a midpoint analysis of the gathered data in order to detect emergent themes or interaction among concepts, which may lead to refinement of the data collection process. Previously collected data may need recoded in light of these refinements; newly collected data can proceed under a revised protocol.

Issues of reliability in ECA are approached differently than QCA. The open-ended format of an ECA protocol precludes the use of standard metrics to assess inter-rater reliability, but efforts can be taken to ensure a level of agreement among coders. For example, as part of the training process, coders should also become familiar with the production process of the medium under study, as well as with the patterns in key examples. Familiarity with the sources facilitates more detailed coding. Research teams employing ECA can promote consistency in coding their observations by coding the same sources independently, then meeting to discuss meanings of concepts and how they are interpreted, as well as recurrent and emergent themes. Coder agreement may be achieved by working together to recode a source. Persistent disagreement may be indicative of another dimension that needs to be in a revised protocol. In general, in ECA, individual consistency in coding is more critical than inter-rater reliability, given the priority placed on reflexivity during data collection.

Analysis and Interpretation in ECA

Data analysis in ECA, like data collection, is a reflexive process. Notes are carefully read, re-read, compared to other notes, sorted, and read again. The early stages of analysis allow the researcher to explore, describe, explain, and perhaps theoretically link elements of the data. In addition, researchers attend to differences within the various conceptual categories. As analysis continues, researchers take notes on their notes, writing small summaries of overarching themes, concepts, and divergent ideas. Summaries include supporting details, such as descriptions of or quotes from the data sources. These smaller summaries are then combined, identifying typical cases and exceptions to those cases, and documenting unexpected elements that push the data in intriguing directions.

Analytical interpretation extends from this process. The small summaries are reviewed in light of the data collection protocol, by which the researcher can return to the original questions of interest. Referring to the protocol allows the researcher to sort the summaries into more distinctive categories to answer those questions, as well as to determine what does not fit and why it does not fit. There may be variation in the nature of the original documents, the concept may have been an unanticipated but important part of the story, or it may suggest a direction for future research. Like QCA, the results of evidence accumulated in ECA are brought to bear on the research questions. ECA, however, is designed not only to tell the story presented in the data but also to reveal, discuss, and contextualize the manifest and latent meaning of that story by grounding it in the social world and the broader social processes by which meaning is produced.

Example of ECA

Welsh, Fleming, and Dowler ( 2011 ) conducted an ECA of crime movies to investigate how crime and victimization were constructed in film, and to discover the messages presented about justice modalities. Given their intention to uncover these themes and messages, they outlined four questions, suggested by the research literature. To select a sample from the population of American films made from 1930–2009 , they performed crime and justice-related keyword searches using the movie reference site imdb.com . From this first set of results, they then employed a theoretical sampling approach, selecting films based on two criteria, derived from their research questions, regarding plot narratives of the films. To collect data, each of the 30 films in the final sample was watched twice. The research questions served as the basis for an open-ended protocol. Coding entailed taking detailed notes about characters’ interactions and about the films’ narratives by observing both dialogue and visual imagery.

Welsh and his coauthors analyzed their data using a constructivist grounded theory approach, which exemplifies the reflexive nature of the ECA process: details from the protocol were read repeatedly and comparatively, with analysis drawing on existing theory. Their analysis revealed three primary themes across the films. In interpreting these three themes, the authors identified several sub-themes, as well as points of contrast. Supporting evidence for each theme is presented as a mix of narrative summary and discussion by the authors and as strategically placed quotations from the films. Where appropriate, connections were made to existing theory, either as points of comparison or as extensions to earlier work. Thus, Welch and his coauthors offered answers to their research questions, supporting their interpretations with descriptive evidence from their sources and with reference to prior literature.

Other examples of ECA include: Altheide and Michalowski’s ( 1999 ) analysis of fear discourse in newspapers; Kort-Butler’s ( 2013 ) study of crime and justice representations in superhero cartoons; and Myers’ ( 2012 ) analysis of televised reports on juvenile detention using program transcripts.

Content Analysis, Crime, and Control

Content analyses of crime and justice issues have covered many genres of media and popular culture. In addition to those examples listed in the sections “Example of QCA” and “Example of ECA,” other genres include, but are not limited to, television crime dramas (Cavender & Deutsh, 2007 ); televised documentaries and realty television (Cecil, 2010 ); television commercials (Maguire, Sandage, & Weatherby, 2000 ); comic books (Phillips & Strobl, 2006 ); music (Hunnicutt & Andrews, 2009 ); criminal justice textbooks (Burns & Katovich, 2006 ); Internet news (Sjøvaag & Stavelin, 2012 ); and movie reviews posted online by imdb.com users (Gosselt, van Hoof, Gent, & Fox, 2015 ).

Across these genres, content analyses in the past few decades have challenged the accuracy of media presentations of crime and justice issues, generally finding that the media often exaggerate the reality of crime by focusing on unusual or rare crimes. Media presentations also tend to misrepresent offenders and victims in ways that do not represent their actual distribution in the population by race, gender, or age, such that young male minorities are seen as the default “typical” criminal. Thus, the nature of crime, as portrayed in the media and popular culture, is violent, the image of the criminal is the cold yet rational predator “other,” and the image of the victim is the innocent prey (Cavender, 2004 ; Kappeler & Potter, 2005 ; Surette, 2014 ).

Content analyses have also revealed the ways in which the media frame stories about crime to correspond to and reinforce these images. Mainstream media depictions of crime and justice generally present messages that conform to and promote the dominant ideology about “the crime problem” and how to manage it (Altheide, 1997 ), namely through established channels of reactive policing and punitive punishments. As that story changes in response to larger cultural shifts in public and political attitudes about crime and control, the methods offered by content analysis are being deployed to understand how the mediated images and meanings about crime and justice will change in turn.

The Relationship between Media and Crime & Media Portrayals of Criminals

Related Papers

Hatice Han Er

Ola Kazeem Falodun

Role of media in relation to crime and deviance has become a very important factor in our society. Communities and individuals are bombarded constantly with messages from different media channels which promote not only products, but moods, attitudes, and a sense of what is and is not important. This essay aims to explore the relationship between media and crime and as well discuss some of the major theoretical positions that focus on this relationship like the media effects theory from both a psychological and sociological perspectives.

IJIP Journal

In this era of technological advancement, media as technologies have dramatically affected the human behaviour and have easily drawn a connection between the virtual world and the real world. Digital technologies are not only the daily tools of communication, information and entertainment for people but have become omnipresent and an integral part of people"s lives. In today"s digital world, the media can give you ideas, it can inspire you to do certain things and drive you to initiate what you see. The media controls the society and is capable of changing human behaviour, living style and moral thoughts and consequences. Media reflects in two ways i.e. positive as well as negative; as media does not produces the real happening of a world, rather creates a new world. The main aim of this research is to study the impact of media on human behaviour. It also aims to create difference between the virtual and the real world. This study would be able to bring out whether trending mass media does influence the youth"s behaviour so as to decide whether to act in a manner parallel to the virtual world or see the practical-real aspect. And also aims to examine the effect of time period of accessing media on the human behaviour. The research design used is "Survey through Questionnaire" with an accidental sampling of 60 people between the age of 15-24 yrs (youth). The questionnaire includes questions pertaining to the subject"s behaviour, influence, preference and thoughts. Everyday incidents relating to recent media trends is visible which leads people to commit acts which have been influenced by popular media and how such behaviour can be disastrous to human life. The newspaper articles shows the changes in human behaviour according to media which is also be studied to lay a base of media"s psychological impact and further interpretation has been based according to the replies given to questionnaire as to what extent it can be influencing, harmful and how should media be utilised. The expectations from this research would be that the behavioural changes are brought by media in the youth and that these changes are quite significant which leads them to act accordingly. Media are considered as the catalyst of change and development, and expected to accelerate the process of social welfare and empowerment of all. The process of communication has created informed citizenry. Playing the role of gatekeeper, media has also contributed in people"s empowerment, diffusion of innovations and ideology of progress and modernity. In a democratic country like India media has been perceived as the friend and voice of the 1 Student Pursuing B.A. LL.B. (Hons.), Currently 4th Year, College

Mass Communication & Society

Kathleen Custers

Asian Journal of Criminology

John N. Erni

Giuseppe Fiorenza

The present work aims to analyze the dynamics of the influence that the media exercise in order to understand how these affect people behaviour and actions, this, focusing on two of the most influencing medias that since last century attracted an always wider quantity of consumers: cinema and television. Special attention will be given to the violent and sexual representations in the American movies. The reason of the choice underlays in the high frequency with which violent and pornographic contents appear on both, the big and the small screen, which makes violence and sex the most influencing themes in media representations. The objective is to prove with a critical analysis taking into account American and European percentages and data over media messages and their consumption, how and why, independently from sex, education level and family influence, the media and the new mediatic communication techniques can actually reshape social dynamics transporting the fictional world in the real one using media texts and psychological techniques as a vector to reach minds and affecting the representation we make and give about reality.

Jordan Cashmore

The fear of crime phenomenon is widely documented and studied in criminology; from the tools that formulate and amplify it to the effects it has on the individual. It is unusual, then, that prior to the last couple of years, few have commented on the positive effects of the fear of crime. More confounding still is the fact that, despite many studies being conducted on the effect of the media on fear of crime, scholars have thus far failed to acknowledge the media’s role in the possibly positive, crime reducing effects of the fear of crime. This paper attempts to make a small step in redeeming this failing in criminology, and proposes a ‘Fear of Crime-Media Feedback Model’, whereby the media influences levels of fear in its audience and affects their routine activity, thereby affecting their exposure to potentially victimising situations by encouraging avoidance behaviour; influencing people to remain in their home as opposed to venturing onto the streets. While the model is currently hypothetical, further research into its plausibility as a crime reducing tool is proposed, potentially providing a method of reducing victimisation risk in the public with use of crime media.

Journal of forensic psychology practice

Christopher Ferguson

Laramie Taylor

Media violence poses a threat to public health inasmuch as it leads to an increase in real-world violence and aggression. Research shows that fictional television and film violence contribute to both a short-term and a long-term increase in aggression and violence in young viewers. Television news violence also contributes to increased violence, principally in the form of imitative suicides and acts of aggression.

Loading Preview

Sorry, preview is currently unavailable. You can download the paper by clicking the button above.

RELATED PAPERS

Lala Jabbarova

Banuka N De Silva

Sara Luísa Meijer Loja

Negative Impacts of Mass Media on Social and Psychological life of Human Beings

Qudratullah Sadiqi

Criminology

Chris Greer

Sofia Khairron

Journal of Applied Social Psychology

Miguel Clemente

Djordje Ignjatović , Milana Ljubicic

David Gauntlett

Mark D Griffiths

Vic Strasburger

David A Green

christie N chukumati

Ryan Tapscott

Çukurova Üniversitesi Eğitim Fakültesi dergisi

Turhan Şengönül

Journal of Communication

Daniel Romer

Canadian Journal of Criminology and Criminal Justice

Rhonda Dobbs

Journal of Contemporary Criminal Justice

Billy Henson

The Oxford Handbook of Criminology

Chris Greer , Robert Reiner

DePaul L. Rev.

Craig Haney

ahmad altaiar

American Psychologist

Craig Anderson

We're Hiring!
Help Center
Find new research papers in:
Health Sciences
Earth Sciences
Cognitive Science
Mathematics
Computer Science
Academia ©2024

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

View all journals
Explore content
About the journal
Publish with us
Sign up for alerts
14 August 2024

Has your paper been used to train an AI model? Almost certainly

Elizabeth Gibney

You can also search for this author in PubMed Google Scholar

You have full access to this article via your institution.

Academic publisher Wiley has sold access to its research papers to firms developing large language models. Credit: Timon Schneider/Alamy

Academic publishers are selling access to research papers to technology firms to train artificial-intelligence (AI) models. Some researchers have reacted with dismay at such deals happening without the consultation of authors. The trend is raising questions about the use of published and sometimes copyrighted work to train the exploding number of AI chatbots in development.

Experts say that, if a research paper hasn’t yet been used to train a large language model (LLM), it probably will be soon. Researchers are exploring technical ways for authors to spot if their content being used.

AI models fed AI-generated data quickly spew nonsense

Last month, it emerged that the UK academic publisher Taylor & Francis, had signed a US$10-million deal with Microsoft, allowing the US technology company to access the publisher’s data to improve its AI systems. And in June, an investor update showed that US publisher Wiley had earned $23 million from allowing an unnamed company to train generative-AI models on its content.

Anything that is available to read online — whether in an open-access repository or not — is “pretty likely” to have been fed into an LLM already, says Lucy Lu Wang, an AI researcher at the University of Washington in Seattle. “And if a paper has already been used as training data in a model, there’s no way to remove that paper after the model has been trained,” she adds.

Massive data sets

LLMs train on huge volumes of data, frequently scraped from the Internet. They derive patterns between the often billions of snippets of language in the training data, known as tokens, that allow them to generate text with uncanny fluency.

Generative-AI models rely on absorbing patterns from these swathes of data to output text, images or computer code. Academic papers are valuable for LLM builders owing to their length and “high information density”, says Stefan Baack, who analyses AI training data sets at the Mozilla Foundation, a global non-profit organization in San Francisco, California that aims to keep the Internet open for all to access.

How does ChatGPT ‘think’? Psychology and neuroscience crack open AI large language models

Training models on a large body of scientific information also give them a much better ability to reason about scientific topics, says Wang, who co-created S2ORC, a data set based on 81.1 million academic papers. The data set was originally developed for text mining — applying analytical techniques to find patterns in data — but has since been used to train LLMs.

The trend of buying high-quality data sets is growing. This year, the Financial Times has offered its content to ChatGPT developer OpenAI in a lucrative deal, as has the online forum Reddit, to Google. And given that scientific publishers probably view the alternative as their work being scraped without an agreement, “I think there will be more of these deals to come,” says Wang.

Information secrets

Some AI developers, such as the Large-scale Artificial Intelligence Network, intentionally keep their data sets open, but many firms developing generative-AI models have kept much of their training data secret, says Baack. “We have no idea what is in there,” he says. Open-source repositories such as arXiv and the scholarly database PubMed of abstracts are thought to be “very popular” sources, he says, although paywalled journal articles probably have their free-to-read abstracts scraped by big technology firms. “They are always on the hunt for that kind of stuff,” he adds.

Proving that an LLM has used any individual paper is difficult, says Yves-Alexandre de Montjoye, a computer scientist at Imperial College London. One way is to prompt the model with an unusual sentence from a text and see whether the output matches the next words in the original. If it does, that is good evidence that the paper is in the training set. But if it doesn’t, that doesn’t mean that the paper wasn’t used — not least because developers can code the LLM to filter responses to ensure they don’t match training data too closely. “It takes a lot for this to work,” he says.

Robo-writers: the rise and risks of language-generating AI

Another method to check whether data are in a training set is known as membership inference attack. This relies on the idea that a model will be more confident about its output when it is seeing something that it has seen before. De Montjoye’s team has developed a version of this, called a copyright trap, for LLMs.

To set the trap, the team generates sentences that look plausible but are nonsense, and hides them in a body of work, for example as white text on a white background or in a field that’s displayed as zero width on a webpage. If an LLM is more ‘surprised’ — a measure known as its perplexity — by an unused control sentence than it is by the one hidden in the text, “that is statistical evidence that the traps were seen before”, he says.

Copyright questions

Even if it were possible to prove that an LLM has been trained on a certain text, it is not clear what happens next. Publishers maintain that, if developers use copyrighted text in training and have not sought a licence, that counts as infringement. But a counter legal argument says that LLMs do not copy anything — they harvest information content from training data, which gets broken up, and use their learning to generate new text.

AI is complicating plagiarism. How should scientists respond?

Litigation might help to resolve this. In an ongoing US copyright case that could be precedent-setting, The New York Times is suing Microsoft and ChatGPT’s developer OpenAI in San Francisco, California. The newspaper accuses the firms of using its journalistic content to train their models without permission.

Many academics are happy to have their work included in LLM training data — especially if the models make them more accurate. “I personally don’t mind if I have a chatbot who writes in the style of me,” says Baack. But he acknowledges that his job is not threatened by LLM outputs in the way that those of other professions, such as artists and writers, are.

Individual scientific authors currently have little power if the publisher of their paper decides to sell access to their copyrighted works. For publicly available articles, there is no established means to apportion credit or know whether a text has been used.

Some researchers, including de Montjoye, are frustrated. “We want LLMs, but we still want something that is fair, and I think we’ve not invented what this looks like yet,” he says.

Nature 632 , 715-716 (2024)

doi: https://doi.org/10.1038/d41586-024-02599-9

Reprints and permissions

Machine learning

Five ways the brain can age: 50,000 scans reveal possible patterns of damage

News 19 AUG 24

Chatbots in science: What can ChatGPT do for you?

Career Column 14 AUG 24

Weather and climate predicted accurately — without using a supercomputer

News & Views 13 AUG 24

Estonians gave their DNA to science — now they’re learning their genetic secrets

News 26 JUN 24

Not all ‘open source’ AI models are actually open: here’s a ranking

News 19 JUN 24

A guide to the Nature Index

Nature Index 05 JUN 24

Principal Investigator Positions at the Institute for Regenerative Biology and Medicine, CIMR

Regenerative Biology and Medicine, including but not limited to disease immunology, ageing, biochemistry of extracellular matrix...

Beijing, China

The Chinese Institutes for Medical Research (CIMR), Beijing

Principal Investigator Positions at the Institute for Molecular and Cellular Therapy, CIMR, Beijing

We're looking for outstanding scientists at all ranks interested in developing novel therapeutics in all disease areas.

2024 Recruitment notice Shenzhen Institute of Synthetic Biology: Shenzhen, China

The wide-ranging expertise drawing from technical, engineering or science professions...

Shenzhen,China

Shenzhen Institute of Synthetic Biology

Qiushi Chair Professor

Distinguished scholars with notable achievements and extensive international influence.

Hangzhou, Zhejiang, China

Zhejiang University

ZJU 100 Young Professor

Promising young scholars who can independently establish and develop a research direction.

Quick links

Explore articles by subject
Guide to authors
Editorial policies

Trump’s False and Misleading Claims about Harris’ Record on Crime

By D'Angelo Gore , Eugene Kiely , Lori Robertson and Robert Farley

Posted on August 15, 2024

Este artículo estará disponible en español en El Tiempo Latino .

In a campaign rally in Atlanta and other venues, former President Donald Trump has tried to paint Vice President Kamala Harris, a former California attorney general and San Francisco district attorney, as being soft on crime. Several of his claims go beyond the facts or distort her positions.

Trump falsely said that Harris supports “mass amnesty and citizenship” for “all illegals … even criminals.” The Trump campaign pointed to past proposals that Harris has supported, but none would have done that.
He wrongly claimed that there was “no crime” if someone stole less than $1,000 worth of goods, and he was wrong again in saying that Harris “started” such a policy. In California, it’s a felony if someone shoplifts more than $950 worth of goods, but it’s a misdemeanor for values below that.
In discussing crimes committed by immigrants living in the U.S. illegally, Trump falsely claimed that Harris, when she was the San Francisco DA, “wouldn’t arrest murderers. She wouldn’t arrest anybody.”
Trump claimed that Harris “supports mandatory gun confiscation” that would leave Americans “defenseless in your house.” Harris hasn’t called for confiscating all guns, and her campaign said she no longer supports the mandatory buyback program for “assault weapons” that she called for during her last presidential campaign.
He claimed that Harris “urged her followers to donate to bail rapists and murderers out of jail” and said “that the violent mobs should not stop.” In June 2020, Harris asked for donations to a bail fund to help people protesting against police brutality, not rapists and murderers. She also said that “protests,” not riots, are “not going to stop” and “should not” stop.
Trump said Harris “supports abolishing cash bail, which means bloodthirsty criminals that just killed somebody can immediately leave custody, go out and kill somebody else.” But ending money bail does not mean accused murderers would automatically be released from jail before trial.
He distorted the facts in claiming that Harris, as California attorney general, “defined and redefined child sex trafficking, assault with a deadly weapon, and rape of a unconscious person … as nonviolent.” A 2016 ballot referendum used the term “nonviolent felonies” without defining them, while the state penal code only specifies 23 crimes as a “violent felony.”
Violent crime and murders nationwide have gone down during the Biden/Harris administration, according to FBI and other crime data. But Trump has taken to citing a different set of data to claim “there’s been a 43% increase in violent crimes since I left office.”

Comprehensive Immigration Reform

In Atlanta on Aug. 3, Trump wrongly claimed that Harris supports “mass amnesty and citizenship” for “all illegals … even criminals.”

Trump, Aug. 3: And the main thing, think of this, she wants to let mass amnesty and citizenship, she wants all illegals to have mass amnesty. Everybody, even criminals.

The Trump campaign referred us to past proposals that Harris has supported, but none of them would have given “mass amnesty and citizenship” to “all illegals,” including criminals.

Specifically, the Trump campaign cited Harris’ support in 2015 as California attorney general for then-President Barack Obama’s Deferred Action for Childhood Arrivals program. DACA defers deportation proceedings for two years, subject to renewal, for qualified individuals — known as Dreamers — who were brought to the U.S. illegally prior to 2007 when they were children.

But these so-called Dreamers are a sliver of the illegal immigrant population, and they are subject to eligibility requirements, including criminal background checks that could result in deportation. As of March 31, there were more than 528,00 active DACA recipients, according to the U.S. Citizenship and Immigration Services.

As a senator, Harris co-sponsored a bipartisan bill introduced by Republican Sen. Lindsey Graham and Democratic Sen. Dick Durbin — the Dream Act of 2017 — that would have made an estimated 2.14 million eligible for conditional legal status and 1.73 million eligible for lawful permanent residence status, according to the Migration Policy Institute. The bill would have required a criminal background check, among other stipulations. As she did as a presidential candidate in 2019, Harris has continued to support “legislation that creates a path for citizenship” for Dreamers as vice president .

The Trump campaign also cited Harris’ support for comprehensive immigration proposals that would provide permanent legal status or a path to citizenship for more than just Dreamers.

The campaign’s “official war room” account on X recently posted an eight-year-old clip of Harris saying she supports “comprehensive immigration reform.” But supporting “comprehensive immigration reform” isn’t the same as wanting “mass amnesty and citizenship” for “all illegals,” including criminals.

In the video clip circulated by the Trump campaign, Harris is shown speaking on the night that she won her U.S. Senate race in California in 2016. “I intend to fight for a state that has the largest number of immigrants, documented and undocumented, of any state in this country, and do everything we can to bring them justice and dignity and fairness under the law and pass comprehensive immigration reform. Bring them out from under the shadows,” she said.

As a senator, Harris did not introduce or co-sponsor a comprehensive immigration bill, according to Congress.gov , but she supported a comprehensive immigration bill as a candidate for the 2020 Democratic presidential nomination.

The Trump campaign also cited Harris’ support for the U.S. Citizenship Act — a comprehensive immigration bill that the Biden-Harris administration sent to Congress in 2021.

The bill — which was introduced in the Senate and House but failed to advance — would have provided a path to citizenship for the 11 million “unauthorized immigrants” that the Department of Homeland Security estimates live in the United States. But certain criminals would be ineligible, and not all eligible immigrants would attain or even seek citizenship. A Pew Research Center study in 2013 found that just 46% of Hispanic immigrants who were eligible to become citizens actually did, for a variety of reasons, such as balking at the English requirement for the citizenship test.

The bill said immigrants had to satisfy “the eligibility requirements set forth in section 245G(b), including all criminal and national security background checks and the payment of all applicable fees.” Section 245G(b) says immigrants are ineligible if they have a felony conviction or three or more misdemeanors. Another section of the bill said that the secretary of Homeland Security shall “determine whether there is any criminal, national security, or other factor that would render the noncitizen ineligible for status under section 245B.”

In its announcement of the bill, the White House said: “The bill allows undocumented individuals to apply for temporary legal status, with the ability to apply for green cards after five years if they pass criminal and national security background checks and pay their taxes.”

Shoplifting

Trump also wrongly claimed at his Atlanta rally that there was “no crime” if someone stole less than $1,000 worth of goods, and he was wrong again in saying that Harris “started” such a policy.

“It used to be that if somebody robbed your store, you would be very tough about it right?” Trump said. “Now, Kamala started this, anything under $1,000, you can do whatever you want. … There’s no crime if it’s less than $1,000. She’s the one who started it.”

He repeated the claim in an Aug. 15 news conference in New Jersey, saying , “You’re allowed to rob a store as long as it’s not more than $950. … They can rob it and not get charged. That was her that did that.”

Trump is referring to a 2014 state law, approved in a referendum by California voters, that reduced the penalties for certain nonviolent property and drug crimes from felonies to misdemeanors. At the time, a 2010 state law had already increased the dollar threshold for a grand theft violation, a felony, from $400 to $950. That meant that thefts of lesser amounts were typically misdemeanors.

The 2014 referendum, called Proposition 47, codified that shoplifting property of $950 or less would always be a misdemeanor, with exceptions for offenders who had prior convictions for certain violent or serious crimes. The measure did not say that thefts under that amount wouldn’t be prosecuted at all. Proposition 47 was proposed at a time of prison overcrowding in California, when the state was under federal court orders to reduce the prison population due to violations of the Constitution’s cruel and unusual punishments clause.

“Prop. 47 created a new category of theft — ‘shoplifting’ (for thefts from retail establishments) — but did not change the dollar threshold” for felony thefts, Michael S. Romano , director of Stanford Law School’s Three Strikes Project and chair of the California Committee on the Revision of the Penal Code, told us in an email.

Harris wasn’t responsible for the 2010 law. That legislation was signed into law by Republican Gov. Arnold Schwarzenegger on Sept. 30, 2010. Harris was the state’s attorney general from January 2011 until January 2017. And we couldn’t find any record of Harris taking a position on Proposition 47, which 60% of voters approved in the November 2014 general election.

Romano told us attorneys general usually don’t take positions on ballot referendums. “Kamala Harris had no role in Prop. 47. In California, the Attorney General has responsibility for writing the summary of all propositions that appear on voter ballots. This is a purely ministerial role,” he said. “Given this responsibility, California Attorneys General rarely, if ever, weigh in on ballot measures. Harris certainly did not take any position on Prop. 47.”

Even if she had, the proposition didn’t say there would be “no crime” if the amount of a theft was “less than $1,000,” as Trump claimed.

It’s worth noting that the threshold for charging property theft as a felony varies by state, but many have set the mark at $1,000 or more, at least for first-time offenders, according to a list of such thresholds, updated as of July, provided to us by the National Conference of State Legislatures. There are several conservative states on that list, including Arkansas , Nebraska and Texas .

The Trump campaign acknowledged to us that Harris didn’t take a public position on the referendum, and it wrongly claimed that “her office misled voters by titling Proposition 47 as the innocuous ‘Safe Neighborhoods and Schools Act,'” pointing to a Spectrum News story in which a Republican state assemblymember made that claim. The supporters of the measure are the ones who called it the “Safe Neighborhoods and Schools Act,” as shown on the December 2013 letter the supporters sent to the attorney general’s office, asking that the attorney general “prepare a circulating title and summary” for the initiative to appear on the ballot.

The official title of Prop 47, written by Harris’ office, was: “Criminal Sentences. Misdemeanor Penalties. Initiative Statute.”

Harris’ office was responsible for the 100-word summary of the proposition that also appeared on the ballot. The Trump campaign pointed to a Sacramento Bee editorial in 2015 that took Harris to task for not mentioning in the summary that Prop 47 could lead to a reduction in DNA samples collected from crime scenes, because more crimes would be misdemeanors. Readers can make up their own minds as to whether that should’ve been in the summary.

But, regardless, it doesn’t show that Harris “started” a policy of “no crime” for shoplifting “anything under $1,000,” which was the false claim that Trump made.

Critics of Prop 47 have backed a new referendum, which will be on the ballot in November, to increase the punishment for some of the crimes Prop 47 affected. Shoplifting goods worth $950 or less would still be a misdemeanor for first-time offenders, under the new Prop 36, but it would be a felony “if the person has two or more past convictions for certain theft crimes (such as shoplifting, burglary, or carjacking),” according to a summary posted by the nonpartisan Legislative Analyst’s Office in California.

Harris as San Francisco’s District Attorney

Harris was San Francisco’s district attorney for eight years, from January 2004 until January 2011. The city, for a period of that time, followed a sanctuary city policy that prevented city officials from turning over minors living in the U.S. illegally to Immigration and Customs Enforcement, even if they had been convicted of a felony.

That policy changed in 2008, with Harris’ support , to allow city law enforcement officials to notify ICE when minors were arrested — and not yet convicted — of a felony crime.

In Atlanta, Trump criticized Harris for shielding “illegal alien, crack dealers from deportation,” and falsely claimed that Harris, when she was the city’s DA, “wouldn’t arrest murderers. She wouldn’t arrest anybody.”

When we asked the Trump campaign for evidence that supported the former president’s statements, we were given links to two news articles from 2008. Both articles contradict Trump’s claim that Harris “wouldn’t arrest murderers” and “wouldn’t arrest anyone.”

The news articles were published during a debate in the city over San Francisco’s sanctuary city ordinance, which dates to 1985. The ordinance initially covered only “immigrants seeking asylum from El Salvador and Guatemala,” but was extended in 1989 to include all immigrants, according to a 2011 article in the UC Davis Journal of Juvenile Law & Policy on the city’s policy.

“At that time, San Francisco officials interpreted the ordinance and state juvenile law as preventing them from referring undocumented immigrants in the juvenile justice system to federal authorities for deportation,” the journal article said.

It is that exception for minors that became an issue in 2008 and that Trump referenced in his Atlanta speech.

The Trump campaign sent us a July 1, 2008, story by the San Francisco Chronicle about eight minors from Honduras who were “convicted of dealing drugs.” The minors were not deported, and they escaped from a group home.

The campaign also referred us to a July 31, 2008 , Chronicle article about a protest against San Francisco’s sanctuary city policy. That article mentioned that Edwin Ramos, a Salvadoran migrant who was living in the U.S. illegally, was “charged with three counts of murder” in a triple homicide. “Ramos, now 21, was convicted for an assault and an attempted robbery when he was 17, but city officials did not turn him over to federal immigration authorities,” the article said. Ramos was found guilty of the murders and sentenced to life in prison.

As those articles show, Trump was wrong. Harris’ office prosecuted and convicted the eight Honduran drug dealers and Ramos.

Those two incidents in 2008 prompted then-Mayor Gavin Newsom to change San Francisco’s sanctuary city policy over the objection of the city’s Board of Supervisors. Newsom directed city officials to report juveniles arrested on felony charges to ICE. In response, the supervisors adopted a new ordinance that allowed the city to report juveniles to immigration officials only after they had been convicted — an ordinance that Newsom vetoed and then ignored after his veto was overridden by the board.

Harris supported Newsom’s policy position, which became an issue for her when she unsuccessfully sought the Democratic presidential nomination in 2020.

As for shielding undocumented minors from deportation, Harris said in 2008 that the city’s “custodial agencies” — specifically the city’s sheriff’s office — were responsible for making referrals to federal immigration officials, not prosecutors.

That appears to be the case. In 2008, Newsom wrote an opinion piece for the Chronicle that indicated the sheriff’s office was responsible for making referrals to ICE.

“Newsom acknowledged that ‘[t]he system failed’ in the case of Edwin Ramos. However, he also emphasized that ‘[o]n a monthly basis, the sheriff refers to ICE some 80 to 100 undocumented individuals arrested for felonies,'” according to the 2011 UC Davis Journal of Juvenile Law & Policy article.

Trump claimed that Harris would make Americans “defenseless in your house” by banning guns.

“She supports mandatory gun confiscation,” Trump said in Atlanta. “Would anybody mind if they came into your house and took away your gun or your guns? … She’s for taking away all of your guns.”

In Montana on Aug. 9, Trump said four times throughout his speech that Harris wanted to “take away your guns,” once saying this would include “all of your guns.”

That’s misleading. While running for president in 2019, Harris said multiple times that she would support a mandatory buyback program for so-called “assault weapons” — but not all firearms.

In a November 2019 NBC News interview , for example, Harris said “there are certain types of weapons that should not be on the streets of a civil society,” referring to assault weapons, which she called “weapons of war.” She said she rejected the idea that “either you’re in favor of the Second Amendment or you want to take everyone’s guns away.”

As of 2024, Harris still supports a ban on purchasing assault weapons. However, her campaign told us that she is no longer advocating that Americans be required to turn in assault weapons that they already own.

Trump distorted the facts by claiming that Harris raised money to bail out “rapists and murders” and that she said people “should not stop” rioting in 2020.

“And during the left-wing riots of 2020, you saw that, she urged her followers to donate to bail rapists and murderers out of jail while saying that the violent mobs should not stop,” Trump said in Atlanta. “And she said, ‘Violent mobs, let the violent mobs keep going.’”

That’s not what Harris did or said.

As we’ve written , days after George Floyd was killed by a police officer in Minneapolis in May 2020, Harris and other well-known figures asked the public to help protesters who had been arrested by donating to the Minnesota Freedom Fund, a nonprofit that pays bail for individuals in jail while their court cases are pending. Harris posted on social media: “If you’re able to, chip in now to the @MNFreedomFund to help post bail for those protesting on the ground in Minnesota.”

The organization said it used some of the money it raised to pay bail and legal expenses for people who were arrested while protesting, but it also covered bail for some people who were jailed for other reasons. However, Harris only asked for donations to help protesters — not “rapists and murderers,” or people charged with committing other crimes.

Harris also did not encourage people to riot after Floyd’s murder. Trump was likely referring to comments that she made in a June 18, 2020, appearance on Stephen Colbert’s CBS talk show. During part of the interview, Harris, who said she participated in protests in Washington, D.C., after Floyd’s death, talked about the role that demonstrations play in getting the government to address social inequities.

After Harris answered his question about the importance of sustaining protests to affect change, Colbert said he had noticed a lack of media coverage of demonstrations that were still happening. That’s when Harris said : “They’re not going to stop. This is a movement, I’m telling you. They’re not going to stop. And everyone beware, because they’re not going to stop. They’re not going to stop before Election Day in November, and they’re not going to stop after Election Day. And everyone should take note of that on both levels. They’re not going to let up, and they should not, and we should not.”

But Harris never mentioned “violent mobs” or encouraged rioters to continue that kind of behavior. She and Colbert only talked about protests, which were overwhelmingly peaceful in the three months following Floyd’s death, according to a September 2020 analysis by the U.S. Crisis Monitor project. Harris even spoke out against violent protests in remarks on Aug. 27, 2020, after demonstrations following the police shooting of Jacob Blake in Kenosha, Wisconsin, led to two deaths .

Ending Cash Bail

Trump accused Harris of backing a plan that allows accused murderers to get out of jail before trial.

“Like other Marxist DAs, Kamala supports abolishing cash bail, which means bloodthirsty criminals that just killed somebody can immediately leave custody, go out and kill somebody else, which, by the way, they often do,” Trump said in Atlanta. He repeated the claim in an Aug. 13 interview on X with the platform’s owner and Tesla CEO, Elon Musk, saying “cashless bail” meant that “you kill somebody and they let you out right away. … And then they never find the people unless they kill again, and then they let them out again.”

In 2019, Harris released a plan “to fundamentally transform our criminal justice system” that said cash bail should be ended. Bail , which is often set by a judge, is the amount of money that a person accused of a crime must pay in order to be released from pre-trial detention while the case is pending.

Harris argued that excessive money bail is unjust and discriminates against “low-income communities and communities of color” who are less likely to be able to afford it. But ending cash bail does not necessarily mean that people charged with murder would be released from custody, as Trump said.

“As you note, it depends on what cash bail is replaced with,” Kellen Funk , a legal historian and law professor at Columbia Law School, told us in an email. “No serious commentator or politician has proposed eliminating cash bail and replacing it only with automatic release.”

Funk said proposals often “replace cash bail with judicial discretion,” which gives judges the option to detain defendants until trial, release them with “nonfinancial conditions” such as drug testing and a curfew, or release them without conditions except a requirement to appear in court.

For “felony homicide,” he said, “the law might recommend detention or set a presumption in favor of it, and in most cases in practice judges are inclined to order detention when the charges involve violent felonies.”

Ames Grawert , senior counsel in the Justice Program at New York University’s Brennan Center for Justice, made a similar point in a phone interview. He told us that policies eliminating cash bail are typically aimed at lower-level offenses, not murder.

“It’s just not the case that bail reform somehow makes it automatic that someone is released, especially if they are accused of a very serious offense,” Grawert said.

Harris’ 2019 proposal mentioned the Pretrial Integrity and Safety Act , which she and Republican Sen. Rand Paul introduced in the Senate in 2017. The bill authorized grants for states to reform or replace their cash bail systems with “individualized, pretrial assessments” that factor in the “risk of flight and risk of anticipated criminal conduct posed by a defendant while on pretrial release.”

Although the bill said there would be a “presumption of release in most cases,” a “preventative detention protocol” could be implemented “for cases in which a judicial officer determines, by clear and convincing evidence and after a hearing during which the defendant is represented by counsel, that the appearance of the defendant in court and the safety of the community cannot reasonably be assured through the use of any combination of [release] conditions.”

In other words, a suspect could be detained if a judge found that it was warranted.

Nonviolent Crimes

In Atlanta, Trump distorted the facts in claiming that Harris, as California attorney general, “defined and redefined child sex trafficking, assault with a deadly weapon, and rape of a unconscious person — an unconscious person raped — as nonviolent. These were nonviolent crimes,” adding that this was done “to help spring child predators and dangerous criminals out of prison all over the state.”

Harris didn’t define those things as “nonviolent.” The California penal code specifies 23 crimes as a “violent felony,” saying they “merit special consideration when imposing a sentence to display society’s condemnation for these extraordinary crimes of violence against the person.” The list includes sexual abuse of a child, rape and “great bodily injury,” as violations of specific laws. That means the list doesn’t cover everything that would be considered violent, and any felonies not specified are technically “nonviolent.”

As support for Trump’s claim, the campaign points to another ballot referendum, Proposition 57 in 2016, which also sought to deal with prison overcrowding by, among other things, allowing people convicted of “nonviolent felonies” to be considered for early-release parole. As the attorney general, Harris was responsible for writing the summary for Prop 57. That summary said : “Allows parole consideration for persons convicted of nonviolent felonies, upon completion of prison term for their primary offense as defined.”

An analysis of the measure by the state Legislative Analyst said : “The measure requires CDCR [California Department of Corrections and Rehabilitation] to adopt regulations to implement these changes. Although the measure and current law do not specify which felony crimes are defined as nonviolent, this analysis assumes a nonviolent felony offense would include any felony offense that is not specifically defined in statute as violent.”

Prop 57 was approved by 64% of voters. CDCR did later issue regulations, saying that cases involving people registering as sex offenders couldn’t be considered by the parole board for early release under Prop 57. But the state Supreme Court struck down those regulations in 2020, saying they violated Prop 57.

That doesn’t mean sex offenders had to be released – it was up to the parole board. But they had to be considered for early release. “We emphasize that this determination does not require the release on parole of any inmate,” the court’s ruling said . “The evaluation of an inmate’s suitability for parole and the processes involved in conducting that evaluation remain squarely within the purview of the Department and the Board of Parole Hearings. … The Department is not permitted, however, to entirely exclude from parole consideration an entire class of inmates when those inmates have been convicted of nonviolent felony offenses.”

At any rate, Harris didn’t “define” or “redefine” the crimes Trump listed.

Nationwide Crime Data

Violent crime and murders nationwide have gone down during President Joe Biden’s term, according to FBI and other crime data, as we’ve written . But Trump has taken to citing a different set of data, the National Crime Victimization Survey.

“Nationwide, there’s been a 43% increase in violent crimes since I left office, including a 58% increase in rape, 89% increase in aggravated assault, and a 56% increase in stone-cold robbery,” Trump said in Atlanta, comparing data from 2020 to 2022 in the government’s National Crime Victimization Survey .

Typically, experts cite crime trends based on data from the FBI’s National Incident-Based Reporting System, which informs the annual Uniform Crime Report. Those are the statistics about crimes recorded by local and state law enforcement agencies around the country, and weighted to compensate for agencies that don’t provide data.

By contrast, the National Crime Victimization Survey provides estimates of crime based on interviews with a sample of people age 12 or older about crimes they experienced in the previous six months. It therefore captures many crimes that are never reported to police.

Experts who track crime in the U.S. say both are useful — albeit flawed — tools to capture the contours of crimes committed in the U.S. But as we said, experts typically refer to the FBI data when reporting crime trends. And, in fact, in past years when criticizing Biden for rising crime in 2021, Trump regularly cited that FBI data. Now that FBI data show violent crime decreasing, Trump has taken to calling those statistics “fake numbers.” They’re not .

“The victimization data and the FBI UCR measure different things,” Richard Berk, emeritus professor of criminology and statistics at the University of Pennsylvania, told us via email. “The victimization data come from surveys in which respondents report if they have been ‘victimized’ by ‘crimes’ asked about in very simple ways. Whether those victimizations were really crimes is hard to determine. And many of those ‘crimes’ are not reported to the police. Also, the willingness to report crime to the police can change rapidly over time depending on such things are media coverage. The UCR is based on crimes reported to the police and depends on voluntary cooperation from departments all over the county. Compliance has long been an issue because some department[s] do not cooperate. Of late, however, compliance has been improving substantially.”

While the two datasets usually follow the same trend lines, experts said , some years they do not. And 2022 was one of those years. The FBI reported that the rate of violent crimes — including murder — decreased in 2022, but the NCVS for 2022 — the latest year available — showed the serious violent crime victimization rate — which includes rape and sexual assault, robbery, and aggravated assault — rose from 5.6 the year before to 9.8 violent crimes per 1,000 population age 12 and older.

In an October report , criminologists at the Council on Criminal Justice wrote that the divergence between the FBI’s Uniform Crime Report and the NCVS “makes it uncertain whether violent crime actually went up or down in 2022.”

Nonetheless, crime analyst Jeff Asher, co-founder of the New Orleans firm AH Datalytics , warned it would be “best to avoid using NCVS to make declarative statements about crime trends,” as Trump did.

For starters, Asher told us via email, “it’s 2024 and the most recent NCVS covers 2022. Reported crime data does a much better job of highlighting current trends than NCVS.”

According to the FBI’s preliminary 2023 crime report , violent crimes dropped 5.7% between 2022 and 2023, and the number of murders declined 13.2%. The official 2023 report will be released this fall.

Asher also warned that NCVS is a survey “and 2020 and 2021 were particularly hard years to survey. The rise in violent crime in NCVS is predicated on violent crime being at historic lows in 2020 and 2021, and I’m skeptical that was the case.”

Asher also notes that NCVS doesn’t include murder — it’s a survey based on interviews with people, which by definition excludes those who have been murdered — “and murder is the crime with the highest societal cost. It’s also the crime that suffers the least from underreporting, so we have a good understanding of the murder trend which is a historic decline in 2023 and 2024.”

“The main weakness of the UCR is that it can only include crimes reported to the police,” Berk concurred. “So, both statistics have flaws, but those flaws differ. The best way to overcome these measurement problems is to focus on crimes that are NOT underreported and quite easily defined. Homicide is a good example because there is a body. A bad example is sexual assault because it is so underreported and can be hard to easily define from survey responses except in extreme cases.”

Data from the Major Cities Chiefs Association, with the addition of New York City’s statistics , show the number of murders has gone down by 9.1% from 2020 to 2023 in 70 large U.S. cities. Data collected by Asher show a 17.6% decline in murders so far this year, compared with the same point in 2023, in more than 250 cities.

And because it is a survey, Asher said, “NCVS has margins of error that add a ton of uncertainty. The rise in violent crime measured in NCVS can vary from an increase in the teens to a 60 to 70 percent increase.”

Finally, Asher said, “NCVS measures from year of survey rather than year of offense which makes picking up suddenly changing crime trends very difficult through NCVS. The survey asks if people have been the victim of a crime in the preceding six months, so it includes crimes from July 2021 to November 2022.”

“Put together, these factors make it particularly challenging to describe the exact shape of crime trends with strong confidence,” Asher said of using only the NCVS.

Berk recommended that people focus on local, rather than national, crime rates.

“Crime is a local issue,” Berk said. “The amount of crime and the kinds of crimes differ enormously in different regions of the country, different urban areas and different neighborhoods within those urban areas. So does citizen cooperation, police resources and police practices. To understand crime and the risks, citizens have to look at the data from their neighborhoods. … All the talk about crime at a national level is statistical nonsense that is nothing more than a political football. And it is a football in a political game over which national level policy can have little to no real effect. Local policies and practices can matter.”

Editor’s note: FactCheck.org does not accept advertising. We rely on grants and individual donations from people like you. Please consider a donation. Credit card donations may be made through our “Donate” page . If you prefer to give by check, send to: FactCheck.org, Annenberg Public Policy Center, 202 S. 36th St., Philadelphia, PA 19104.

Grab your spot at the free arXiv Accessibility Forum

Help | Advanced Search

Computer Science > Computation and Language

Title: mutual reasoning makes smaller llms stronger problem-solvers.

Abstract: This paper introduces rStar, a self-play mutual reasoning approach that significantly improves reasoning capabilities of small language models (SLMs) without fine-tuning or superior models. rStar decouples reasoning into a self-play mutual generation-discrimination process. First, a target SLM augments the Monte Carlo Tree Search (MCTS) with a rich set of human-like reasoning actions to construct higher quality reasoning trajectories. Next, another SLM, with capabilities similar to the target SLM, acts as a discriminator to verify each trajectory generated by the target SLM. The mutually agreed reasoning trajectories are considered mutual consistent, thus are more likely to be correct. Extensive experiments across five SLMs demonstrate rStar can effectively solve diverse reasoning problems, including GSM8K, GSM-Hard, MATH, SVAMP, and StrategyQA. Remarkably, rStar boosts GSM8K accuracy from 12.51% to 63.91% for LLaMA2-7B, from 36.46% to 81.88% for Mistral-7B, from 74.53% to 91.13% for LLaMA3-8B-Instruct. Code will be available at this https URL .

Subjects:	Computation and Language (cs.CL)
Cite as:	[cs.CL]
	(or [cs.CL] for this version)
	Focus to learn more arXiv-issued DOI via DataCite

Submission history

Access paper:.

HTML (experimental)
Other Formats

References & Citations

Google Scholar
Semantic Scholar

BibTeX formatted citation

Bibliographic and Citation Tools

Code, data and media associated with this article, recommenders and search tools.

Institution

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs .

IMAGES

CRIM1101 Research Essay
CMC 3
Media & Crime: Contemporary Issues
7 MEDIA AND CRIME POWERPOINT
(PDF) Media made criminality: The representation of crime in the mass media
Paper Presentation On Cyber Crime

COMMENTS

Violence, Media Effects, and Criminology
Scholar Sheila Brown traced how research in criminology evolved from a focus on "media and crime" to what she calls the "media/crime nexus" that recognizes that "media experience is real experience" (Brown, 2011, p. 413). In other words, many criminologists began to reject as fallacy what social media theorist Nathan Jurgenson ...
Media Representations of Crime and Criminal Justice
Along with a host of other topics, matters of crime and criminal justice are made collectively visible by the media (Thompson, 1995, 2011), a process of some significance given that most instances of crime and many actions of criminal justice agencies seek some sort of secrecy. Interesting questions relate to the processes of revelation and ...
Media Criminology
Media criminology is a signific ant area of study that examines the intersection of crime, media, and society. The role that the media plays in constructing im ages of crime, criminals, and ...
Understanding the role of media in the formation of public sentiment
On several occasions, research has demonstrated that the amount of media coverage of criminal events is largely uncorrelated with the actual crime rates, thereby creating the perception that crime ...
Crime and its fear in social media
Social media posts incorporate real-time information that has, elsewhere, been exploited to predict social trends. This paper considers whether such information can be useful in relation to crime ...
Crime News Consumption and Fear of Violence: The Role of Traditional
Crime has been one of the main topics in news media for decades (e.g., Dominick, 1978; Graber, 1980; Johnson, 1998).In the United States, a study by Klite et al. (1997) revealed that more than 70% of the 100 news channels examined in the study opened with a crime-related news story. Similar trends are evident in number of other Western societies, for example, in the United Kingdom, the ...
Media made criminality: The representation of crime in the mass media
studies of media e ﬀects estimated that overall they showed that a single exposure to. media made criminality . violent or stereoty ped cont ent was followed by about an extra 20 per cent of ...
The Role of the Media in the Fear of Crime: A Qualitative Study in the
The present study, using a Portuguese sample, is aimed at understanding how the media shapes the fear of crime experiences. Concretely, we explored how individuals interpret the content transmitted by the media and how these perceptions influence the fear of crime contextually situated. It was also our aim to explore if the area of residence (rural and urban), as well as certain ...
Framing and Cultivating the Story of Crime: The Effects of Media Use
Lisa A. Kort-Butler is an associate professor in the Department of Sociology at the University of Nebraska-Lincoln. She earned her PhD from North Carolina State University. Her research interests focus on media representations of crime and justice and their implications for public opinion.
The Relationship between Media Consumption and Fear of Crime: A ...
Abstract. This review comprehensively analyses the relationship between media consumption and fear of crime. Most reviewed studies indicate a positive correlation between media consumption and fear of crime, which can be attributed to the media's tendency to focus on sensationalised and dramatic crimes and the media's effect on individuals' perceptions of reality; however, several studies have ...
PDF Media Effects on Crime and Crime Style
Media Effects on Crime and Crime StyleM. sity Draft: April 1st, 2018AbstractEvidence about the relationship between exposure to media violen. e and criminal activity remains mixed. While some scholars argue that exposure to violent media contents "triggers" crime and aggression, others contend that media may influence crime, but only as a ...
PDF Crime and media: understanding the connections
Researching crime and media Research on crime and media covers three principal areas of interest: content; production; and consumption and influence. Each area has its own particular research methods and approaches (Greer, 2010). Media content analysis can be split between studies which are primarily either quantitative or qualitative.
How Crime-Based Media Affect Perceptions of Crime, Race, and Fear of Crime
Thus, exposure to either media source could influence perceptions based on crime, race, and fear of crime (Moore, 2011). Individual exposure to crime-based media content. is important to consider in this study. The use of social cognitive theory has been persuasive concerning the criminal.
News media and crime perceptions: Evidence from a natural experiment
Some of the other relevant papers are Jensen and Oster (2009); Chong et al. (2012); Della Vigna et al. (2014); Rizzica and Tonello (2015); and Lim et al. (2015). 9 Our paper belongs to this growing research cluster and it complements it by providing a new field of evidence about the impact of media: the one on crime perceptions.
Content Analysis in the Study of Crime, Media, and Popular Culture
Defining Content Analysis. As a research method, content analysis exists somewhere between purely quantitative and purely qualitative. In the study of crime in the media, research ranges from studies that count or otherwise quantify texts for the purpose of statistical analysis to studies that explore presentation and representation of crime-related issues.
A social network of crime: A review of the use of social networks for
Existing survey papers. Crime and social media is a popular research area, consequently there are a number of survey papers that address this area. However, these survey papers tend to be targeted to a specific crime such as hate speech [1] and Spear Phishing [2], or the use of social media for policing in general [3]. To date, there has not ...
(PDF) Crime and Social Media
development consequences of social media (World Bank, 2016). In response to above policy concern, the positioning of this study on the relationship. between social media and crime is motivated by ...
PDF The Influence of Media on Public Perception of Crime
• The media, crime prevention and urban safety- A brief discussion on media influence and areas of further exploration by Viven Carli: This paper talks about the impact of media on crime prevention by providing and integrative approach for exploring the multiple roles of media. Media holds an important position for communication, and
Crime, Media, Culture: Sage Journals
SUBMIT PAPER. Crime, Media, Culture is a peer reviewed, international journal providing a vehicle for scholars working at the intersections of criminological and cultural inquiry. It promotes a broad cross-disciplinary understanding of the relationship between crime, criminal justice, media and culture. ... Sage Research Methods Supercharging ...
(PDF) The Relationship between Media and Crime & Media Portrayals of
More confounding still is the fact that, despite many studies being conducted on the effect of the media on fear of crime, scholars have thus far failed to acknowledge the media's role in the possibly positive, crime reducing effects of the fear of crime. This paper attempts to make a small step in redeeming this failing in criminology, and ...
[2408.07009] Imagen 3
We introduce Imagen 3, a latent diffusion model that generates high quality images from text prompts. We describe our quality and responsibility evaluations. Imagen 3 is preferred over other state-of-the-art (SOTA) models at the time of evaluation. In addition, we discuss issues around safety and representation, as well as methods we used to minimize the potential harm of our models.
Has your paper been used to train an AI model? Almost certainly
Experts say that, if a research paper hasn't yet been used to train a large language model (LLM), it probably will be soon. Researchers are exploring technical ways for authors to spot if their ...
Trump's False and Misleading Claims about Harris' Record on Crime
Violent crime and murders nationwide have gone down during President Joe Biden's term, according to FBI and other crime data, as we've written. But Trump has taken to citing a different set of ...
Shafik's Columbia: 13 months and 13 days of a campus spiraling into crisis
The Manhattan district attorney initially charged the incident as a hate crime but moved to drop the case in late June. Oct. 12, 2023 Hundreds of protesters packed South Lawn, with the Columbia chapters of Students for Justice in Palestine and Jewish Voice for Peace demonstrating opposite to Students Supporting Israel at Columbia.
NCCD Research Review : Crime and the Mass Media: a Selective Review of
In general, the available research shows that crime is a relatively staple topic in the media's news and entertainment presentations. The amount of crime depicted has little relationship to the amount of crime occurring, and violent crimes are high ly overrepresented by the media.
Bureau of Crime Statistics and Research
The involvement of young people aged 10 to 13 years in the NSW criminal justice system. Link to report summary: The involvement of young people aged 10 to 13 years in the NSW criminal justice system Release date: Wednesday 14 August 2024 In NSW, the minimum age of criminal responsibility is 10 years of age which means that a child aged under 10 years cannot be found guilty of a criminal offence.
(PDF) How Social Media Influence Crimes
Social media's anonymity and fakeness, as well as jurisdictional issues, have contributed to an increase in online violence that affects people of all ages and is a global issue. Individuals ...
Mutual Reasoning Makes Smaller LLMs Stronger Problem-Solvers
This paper introduces rStar, a self-play mutual reasoning approach that significantly improves reasoning capabilities of small language models (SLMs) without fine-tuning or superior models. rStar decouples reasoning into a self-play mutual generation-discrimination process. First, a target SLM augments the Monte Carlo Tree Search (MCTS) with a rich set of human-like reasoning actions to ...
PDF 2023 1internet Crime Report
The past year, investment fraud was once again the costliest type of crime tracked by IC3. Losses to investment scams rose from $3.31 billion in 2022 to $4.57 billion in 2023—a 38% increase. The second-costliest type of crime was business e-mail compromise (BEC), with 21,489 complaints amounting to $2.9 billion in reported losses.
The Relationship Between Social Media Data and Crime Rates in the
These guidelines on ethical conduct should be based on current evidence and standardized while consideration of various laws and practices in different jurisdictions. Future research needs to further explore the ethical issues in social media monitoring research on people's attitude and willingness to share social media data for crime prevention.