Projects 2022-2023

An analysis of punitive attitudes in social media for traditional and cyber-enabled crime.

This article reproduces the article made with homologous data for Uruguay in “Razones y Personas” (Ezquerra and Trajtenberg 2023).

Citizens' attitudes to crime and punishment significantly impact the functioning of the criminal justice system and crime prevention policies. Traditionally, such research has relied on data from public opinion surveys, which is limited by their high cost and cross-sectional nature. Funded by the seed corn funding of the Centre for Digital Trust and Society at the University of Manchester this article delves into the potential of utilising New and Emerging Forms of Data (NEFD) to understand punitive attitudes in social media. Collecting crime-related news stories from the mainstream media in the UK and analysing users' responses, we show that the discussion about punishment and punitiveness occupies a significant proportion of the conversations surrounding crime.

Introduction

For some time now, Social Sciences have been studying punitive attitudes, which refer to perceptions of what kind of punishment we believe is legitimate and valid to apply when a crime occurs. Research has shown that citizens' opinions and preferences for punishment carry serious consequences. For example, they can influence decisions made by politicians and governments regarding the design and implementation of crime prevention policies that are ineffective or even counterproductive (Enns 2014; Jennings et al. 2017) and even highly costly in economic terms (Jaitman et al. 2017; Welsh 2018).

In the UK, even more than 50 years after the abolition of the death penalty, it continues to be supported by 40% of the population (Kirk 2022a). When considering specific crimes (murder of a child, acts of terrorism, or multiple murders), the death penalty is supported by most of the population (ibid). Furthermore, approximately two-thirds of Britons believe that the sentences received by offenders convicted of crimes are not sufficiently severe (Kirk 2022b). This prevailing public opinion regarding punishment takes place in a context of growing economic uncertainty (KPMG 2022) and an increase in violence (pre-pandemic) that had been steadily declining since the mid-1990s (Office for National Statistics, 2021; Office for National Statistics, 2022). Furthermore, the recent covid-19 pandemic and subsequent lockdown measures affected social interactions and criminal strategies in particular, multiplying cyber-dependent and cyber-enabled crime (Buil-Gil et al. 2021), raising the question about the effects of this type of crime on citizens' attitudes.

But what are these punitive attitudes and how can we gain insights about them? Over the past three or four decades, criminologists have tried to address these two questions by relying on conventional methods generally based on scales and surveys. What are the limits associated with using these data sources? What other alternatives are available, and what possibilities and limitations do they present? This article aims to provide a brief overview of these non-traditional data alternatives and the so-called computational criminology field. Along the way, it aims to showcase some initial empirical findings of the project we developed in partnership between the University of Manchester and Cardiff University.

"Hang him, then deport the whole family"

After recognising the importance of citizens' punitive attitudes, a key question is how to observe them effectively. Initially, studies only included very general measures of attitudes towards sentencing and punishment, usually based on a single survey question with minimal contextual information (for example referring to support for the death penalty or the general increase in sentences). This type of design tends to overestimate the punitive attitudes of individuals, as it fails to provide information regarding other types of measures (such as rehabilitative or restorative ones) and does not consider the characteristics of the crime, the victim or the offender (Stalans, 2013).

Over time, measurement strategies became more sophisticated. For example, the use of vignettes has been increasingly adopted in the field, allowing not only to control the characteristics of the crime, but also to manipulate them within experimental contexts (Socia et al. 2019; Horstman et al. 2021). Also, batteries of questions have been applied to enhance the validity of the measures of punitive attitudes through more sophisticated psychometric analysis (Maguire and Johnson 2015; Armborst 2017). Despite these advances, measuring punitive attitudes in this way faces some challenges.

The first problem with these alternatives is the limited flexibility of scales applied in surveys which often fail to capture individuals' opinions, moods and emotional reactions towards punishment validly and reliably (Smith et al. 2017). When respondents are required to express their opinion on sensitive or controversial topics like punishment, there is a social desirability bias and in many cases the response is tailored to what the interviewer or the community expects (Stephens-Davidowitz 2018). The second problem is that these surveys involve, at best, a static and rather costly to implement snapshot that cannot capture the real-time dynamics and volatility of citizen opinion over weeks and months (Williams et al. 2020). Finally, the third problem is that survey-based measures of punitiveness are focused on atomised individuals and thus hardly capture existing networks of communication, exchange and discussion (Edelmann et al. 2020).

So, what is the solution? This is where computational criminology steps in to complement and tackle the limitations of these conventional measures through the use of New and Emerging Forms of Data (NEFD). What exactly are these NEFDs? Basically, it is a grandiose name for multimedia data digitally captured through various types of digital connections (Radanliev and De Roure 2023). The expansion of personal Internet usage and its application to multiple kinds of processes changed how individuals inform and communicate, generating an exponential growth of available data on online behaviours that greatly interest social sciences (Edelmann et al. 2020; Williams et al. 2020). NEFDs provide detailed information on the development of social relationships in large populations as they occur, unlike traditional datasets typically collected by social scientists (Hofman et al. 2021; Radanliev and De Roure 2023). However, it is only recently that criminologists have begun to exploit these data and start studying phenomena such as social disorder, policing, online crime, or racial hate crimes (O’Brien et al. 2015; Smith et al. 2017; Lynch 2018; Williams et al. 2020). As far as we know, no one has ventured into investigating punitive attitudes in this way. In this blog note, we want to outline the potential of this type of data and the associated techniques drawing on the initial outcomes of the study we are carrying out in the United Kingdom.

The black box

In order to analyse punitive attitudes on Twitter, the following data was collected: First, 20 news portals with a presence on Twitter were selected based on their relevance according to public opinion polls and number of followers. From these 20 news portals, all tweets related to crime were downloaded through a set of keywords (e.g., "homicide," "assault," "robbery," "robbery," etc.) through the year 2022. These keywords were derived from exploratory analysis of news media. All quotes or responses made by users to these news items were downloaded together with the originals tweets. This latter dataset holds relevance for our analysis, as it groups user reactions to crime news. This dataset is anticipated to yield punitive reactions, which, in one way or another, call for an increase in the cost of crime (i.e., that committing a crime should have higher costs for offenders through harsher penalties).

After acquiring this data and following the removal of spam and bots, the first step involves attempting to summarise this information. Qualitative analysis of every tweet would be a daunting task, and here is where Topic Modelling strategies prove invaluable. But what exactly is Topic Modelling? Topic Modelling is an analysis technique in the Machine Learning (or Artificial Intelligence - AI) toolkit. This analysis technique is based on a series of sequential steps or instructions where the machine will try to optimise a specific task by adjusting values within a statistical model based on the variation in an indicator which is iteratively employed as a quality criterion (akin to what can be called an "algorithm" in the context of data science). More specifically, topic modelling constitutes a non-supervised learning technique. Unlike in a supervised classification task, we are not giving examples to the algorithm regarding the expected outcome, - but instead, we are giving it data and a goal. In this case, the purpose is categorising (or clustering) the textual information. How is this accomplished? By identifying words that commonly co-occur within documents (in this context, tweets) and that do not tend to occur in other contexts (Blei et al. 2003; Blei 2012).

Why are we doing this? By employing this approach we can accomplish the coding of our documents of interest in an automatic and inductive manner (Mohr and Bogdanov 2013). By detecting these words that distinguish discussion topics, we can identify the main themes of naturally occurring communications within social media, specifically in the "red chronicle" context. Thus, we can answer how crime is discussed in social media, especially to what extent punishment and punitiveness are emerging concerns. All this analysis, in addition, is executed through a beautiful, replicable, automated, and cost-free software package.

If all this seems confusing, don't abandon us (yet): many of these points will probably become clearer with analysis we present next.

• Read the full blog post in Word doc format, including full references and charts

Introduction

Discussions around incels – involuntary celibates – have become prevalent across popular social media platforms such as X/Twitter. However, it is not always clear who exactly is behind the incel identity. Are incels just people who can’t find love, or is there more to it? Can anyone fall into inceldom, or are there any special requirements? What keeps incels hooked on online incel forums? And how can incels balance their online and offline life? I argue incels experience an in-between identity as they pave their way through the online shit-posting and offline sexual frustration. I focus on the current most popular incel platform: incels.is, an incel forum filled with misogyny, homophobia, and racism, established in 2017.

Splitting identities

Before discussing the identity of an incel, it is worth examining what the drivers are behind these identities. In our case, incels.is plays a big role.

The key element that determines the success of incels.is is the anonymity of its users. This is because anonymity gives users the power to establish themselves in a certain way online, without having to endure any negative consequences offline. It is important to point out that because incel-forum users are part of communities which are virtual (incels.is) and physical (universities, workplaces, neighbourhoods, etc.), they will similarly have two identities – also physical & virtual – and will interpret the term ‘incel’ differently across these two identities.

In their physical communities, involuntary celibates understand that being an ‘incel’ has a pejorative meaning, since lacking sexual experiences infringes on their ability to establish themselves as ‘masculine’. Indeed, young men who are virgin-shamed acquire a lower perceived social status, in accordance with ideas of male dominance.^[1] Not being involved in sexual relationships implies a crucial lack in one’s manhood, and a failure to fully transition from child to adult – and subsequently a (self-)denial of desirable masculine identities. Therefore, young men do not want to be associated with lacking sexual experiences offline, or therefore being seen as ‘less masculine’, and attempt to stay clear of publicly self-labelling as ‘incel’.

Lack of recognition

Because young men don’t want to identify as ‘incel’ offline, speaking of sexual discontent appears to be taboo. This fuels the incel’s frustration, and significantly decreases any received recognition associated with their celibacy. I distinguish two categories in which incels lack recognition in their physical communities: intimate and social.

Regarding the intimate category, incels experience not being involved in sexual or romantic relations while other men are as an injustice. They feel they are owed sex from women, since there exist other men who do ‘receive’ sex. This deficiency is felt as an injustice, in the sense of unfairness. Subsequently, they experience a frustration of being denied these ‘female goods’. They therefore desire recognition, or acknowledgement, that they are lacking sex and are being treated unfairly. However, speaking of this would seem humiliating and would appear to entail a lower social status offline – presuming their predisposition to ideas around men’s dominant position in society and the secondary status of women (or in other words: hegemonic masculinity).

Regarding the social category, because incels cannot boast about having had sexual experiences, they cannot gain higher social status within their physical communities. Moreover, they feel they are unable to relate to their peers, make meaningful relationships through boasting about their sexual relations, or have success in their work-life – again, all of which are assumed through ideas of hegemonic masculinity. Hence, they lack recognition because they cannot climb the social ladder, and their frustration around this remains unacknowledged. For these reasons, incels will seek recognition elsewhere – and this is where incel forums come in.

The incel forum

As discussed above, on the one hand, involuntary celibates do not want their social identity to be dominated by their inceldom, yet still desire to be recognised for the ‘suffering’ they must endure. Incel forums therefore provide the most appropriate setting for incels to share their frustration and be recognised as inferior and sex-lacking beings with a common enemy: women.

The anonymity of the forum means incels can be truthful about their lack of sexual experiences, and in this way establish themselves as true involuntary celibates. Since every incel on the forum is in the same boat, sex cannot be used to boost status. Moreover, this entails that fellow incels can appropriately recognise and reciprocate feelings of sexual and social frustration, meaning the incel-forum is also a place to find solace. However, over time incels.is has turned extremely misogynistic, racist, and homophobic, even encouraging fellow members to attack those that are at the root of their inceldom: women and alpha men (who allegedly take all the women).

The forum’s aspect of anonymity means that a member’s identity is not wholly that of an involuntary celibate. Online they can proudly discuss their celibacy and show their hatred of women, but when they shut down their computers their inceldom and misogynistic posts remain unknown by their physical community members. Hence, the latter cannot necessarily perceive them as being true involuntary celibates, meaning incels in this way prevent their identity being fully branded as ‘incel’. For this reason, incels can both self-assert and self-deny their inceldom through their virtual-incel and physical communities respectively: online they can exercise any desired hatred towards women, while their posts do not have any ramifications in their offline lives.

In this way, incels are constantly in-between identities because they experience both a desire and refusal to self-identify as incel. In their physical communities this refusal is most visible in their unwillingness to publicly admit they do not regularly engage in sexual activities. This is done out of a fear of appearing less masculine, since this could further lower their self-perceived social, professional, or sexual status. Their desire to self-identify as incel in their physical communities can be seen in their longing for recognition in their sexual and social frustration, and in their secret joining of the online incel forum.

Similarly, in their virtual incel communities, their desire to self-identify as incel is seen through their engagement with misogynistic, racist, and homophobic posts. Their refusal to self-identify as incel, however, is seen in their online anonymity: in their refusal to upload any personal information onto their online profiles, as this would wholly solidify their incel nature.

This is because, being non-anonymous, any received incel-related recognition would be directed towards their public image (rather than merely their anonymous, online agency), in this way forcing the incel to wholly self-objectify as incel. Such a complete self-identification with the incel-identity would appear undesirable due to its potential, personal risks.

After all, frequently engaging with racist, homophobic, and misogynistic content would (hopefully!) appear unappealing to any employer – meaning loss of one’s job would soon follow. Hence, this continuous desire and refusal forces the incel to be in-between two identities: that of the online frustrated incel, and offline ordinary fellow.

Since involuntary celibates feel they are unable to progress on the social ladders of their physical communities, they can substitute their desire for higher social status in physical communities with that of virtual incel forums. Regular forum posters gain stars, and subsequently increase ranks. Users with a high rank-colour are therefore perceived as having higher trustworthiness and social status compared to e.g., “graycels” – a pejorative reference to users with a post-count below 500, who have a grey rank-colour. Mechanisms such as these create an understanding of acquirable status within virtual incel forums.

Conclusion

Incels find themselves in-between identities while trying to navigate their frustration and anger towards women. Though they cannot express their inceldom offline, they find solace and recognition when expressing it online.

Understanding the mechanisms that underlie the incel’s adoption of identities is key to solving how such damaging mechanisms can be prevented. With increasing terrorist activity linked to incel forums, it therefore evidently remains a very important area of research.

This blog post has been made possible by The Centre for Digital Trust and Society, under the supervision of Mihaela Popa-Wyatt.

• Read this blog post in Word doc format

This post is also available on the Cardiff University Open for Debate blog.

Defining the acceptability of ‘safe’ data linkage to identify women at risk of postnatal complication in Greater Manchester

Dr Victoria Palin, Epidemiologist and Lecturer in Maternal and Fetal health, The University of Manchester

The Seedcorn funding (awarded November 2022) from the Centre for Digital Trust and Security enabled a thoroughly insightful and meaningful piece of work to be conducted within the local community of Greater Manchester.

The problem

Currently there is a disconnect between primary and secondary care services. Sometimes important information, for example when a patient is discharged from hospital after giving birth, this information does not always reach a GP surgery in a timely manner, or at all.

A maternal postpartum 6-week check with a GP is recommended by national guidelines to assess mothers’ recovery from pregnancy and birth, and to evaluate physical and mental health needs going forward. Its estimated 62% of women receive health checks within an extended 12-week period, with just 40% occurring in the recommended 6–8-weeks. It is also shown that women from more deprived areas are less likely to have a check at all, showing disparity in the delivery of clinical care for women after having a baby.

This is particularly concerning for mums that had blood pressure complication before or during pregnancy as they are more likely to develop ischemic heart disease, stroke, and heart failure in later life. To prevent long term health risks to mum (i.e., within the next 5 to 10 years), it is vital mums are seen within a few weeks after birth by a GP and so that their blood pressure health can be regularly managed to reduce the risk of developing long-term complications. Currently, GPs struggle with the limited data flow from hospitals to know who needs a six-week check. Clearly there is a need to automate the way GPs can be notified of women who have recently given birth and now require an appointment, especially those with an increased risk of complication and long-term poorer health.

The solution

Patient electronic health records (EHRs) and digital interventions to optimise care have become a large part of healthcare research. Given the sensitivity of health records, data protection, security, and trust are of uttermost importance. The Greater Manchester Care Record (GMCR) links data from different services to allow clinical teams to quickly access a personalised record of the patient in front of them to support direct care. In collaboration with Health Innovation Manchester and the University of Manchester Research Operations Group, an infrastructure to allow secondary use of data for research is being developed. This infrastructure may be used to safely and securely link, analyse, and identify mothers in the postpartum period with a greater risk of complication, who therefore require early medical intervention.

The work

The seedcorn finding allowed us to successfully interview a diverse group of public participants who reside within Greater Manchester, as well as clinical staff who work across the region. We used these workshops to explore the acceptability of:

• Linking EHRs between maternity services for research to better identify mothers in the postpartum period,
• Using de-identified versions of this data (secondary use) within a secure environment to develop mathematical algorithms that predict individual risk for developing postnatal disease,
• Developing a dashboard to notify primary care doctors which of their patients recently delivered, requiring medical review,
• Prioritising GP review by risk using the mathematical algorithm.

The impact

Public engagement workshops, including 27 contributors of mixed ethnicity, along with clinical interviews of 7 stakeholders ranging from GPs, midwives and health psychologists were conducted. Following a thematic analysis looking at both the acceptability and trustworthiness of secondary use of data in maternity research we are about to publish our best practise recommendations in a high impact peer review journal.

Furthermore, we are now in the process of co-developing an innovative digital solution to improve postnatal care within Greater Manchester that adheres to patient and clinical concerns of safe and secure secondary use of data in research. By incorporating clinical and public opinions at each stage of our work we aim to improve inequalities in accessing healthcare for maternity services, whilst also targeting care to women who need it the most, helping to reduce their risks of long-term health complications.

• Read this blog post in Word doc format, including project pictures

Modern slavery and digitisation in fast fashion supply networks: The transparency dividend

Authors: Jon Davies, Rose Broad, Amy Benstead, Michelle Corallo

Overview

Practices associated with various types of ‘modern slavery’ and exploitation have long been associated with sectors such as garment manufacturing, which forms part of the apparel sector. It has been less clear in practice what role digital technology such as Blockchain and AI tools have in preventing and addressing modern slavery in this sector.

The purpose of this project has been to shed light on the role and use of digital technology by conducting interviews with industry and regulatory stakeholders, as well as an assessment of 89 modern slavery transparency statements from the Fashion Transparency Index Report 2022. Here we outline some context to the topic area, as well as key preliminary findings that suggest a general lack of awareness or lack of commitment from businesses to utilise digital technology in developing their transparency efforts.

Technology, garments, and modern slavery

The fashion industry faces numerous challenges that many consumers can associate with in some way. In addition, the industry is poised for transformation with upcoming Digital Product Passport EU legislation under development. With this in mind, companies are going to have to achieve greater transparency in their supply chains. Whether referring to labour conditions, supply chain complexity, environmental sustainability, counterfeit products, consumer trends and preferences, as well as regulatory compliance, the evolving nature of digital technology has the potential to crosscut these challenges and inform the ways in which businesses and the wider industry develop their commitments to ethical practice and transparency.

For example, Blockchain has emerged as a prime example of how businesses can use real-time, ledger-type tools across supply networks to monitor transactions and ‘smart contracts’. The supposed key benefits include greater transparency due to other businesses being able to view transactions across a supply network; increased efficiency since there is less room for disputes and less need for paper records; and the fact that information cannot be retroactively altered by just one organisation.

In parallel developments, exploitation associated with modern slavery such as forced labour and human trafficking has been gradually shifting up the political and public agenda over the last 20+ years. Common concerns related to poor work conditions within garment manufacturing include under-payment (or non-payment) of salary, excessive working hours, poor health and safety practices, and non-standard employment practices such as zero-hours and/or subcontracted labour.

Whatever the reality in practice, companies have increasingly been keen to talk-up their efforts to improve transparency within their organisations and supply networks, especially since the UK introduced the Modern Slavery Act 2015, which requires larger businesses to publish a yearly statement outlining efforts they have made to address exploitation. However, the use of digital technology such as Blockchain in these efforts seems to be underdeveloped to say the least. This project has revealed that for many organisations in the garment sector, the underuse of technologies may lie in a limited awareness of technological tools and/or a lack of commitment amongst some businesses to tackling social challenges.

Limited awareness of technological tools

From both our interview data and secondary analysis of transparency statements, a recurring theme was that stakeholders connected to garment manufacturing have a limited or very general understanding of digital tools like Blockchain, and how these could be used in a beneficial way, i.e., to contribute to their business activities and efforts to address labour exploitation. Tech companies were clearly more aware and had various ideas of how Blockchain and other tools such as apps could be used within the industry, but similarly felt that a lack of awareness from some companies was hindering this process.

From our sample of 89 modern slavery statements based on apparel companies operating in the UK, 69 do not mention digital technology at all, and most of those that did refer to technology in a superficial way. Our participants felt that many businesses operating on narrow profit margins or without readily available technical expertise would be less able to prioritise inclusion of tools such as Blockchain; or that in some cases, well-established companies relied more on “gentlemen’s agreements” as opposed to verifying information and data about business transactions. To some extent such attitudes fed into an overall perceived lack of commitment from some businesses to take concerns over exploitation and associated matters seriously.

Lack of business commitment

Due to limitations on resources in many cases, it would be unfair to brand all garment manufacturing companies not engaging with tools like Blockchain as showing a ‘lack of commitment’ to tackle social challenges within their businesses and supply networks. That said, an example of how companies do engage with digital technology is that of H&M, who state that in 2022 they launched several pilots with ‘TextileGenesis’, which uses Blockchain technology to monitor textiles through the supply network.

Even so, there was a broader perception from stakeholder discussions that some businesses are more interested in paying lip service to the idea of transparency rather than examining underlying factors such as business and employment models that contribute to exploitation. This may be partially because the penalties for non-compliance in the UK are not regarded as sufficiently severe to ensure the positive development of good practice.

Related to this, AI tools that monitor companies’ transparency commitments (statements) can be traced over a period of years, thereby assessing whether, for example, companies have rowed back on commitments previously made or cut out significant detail that may have previously been a key highlight or goal. However, this is a volatile issue to monitor, and is reflected by the fact that our assessment of transparency statements from the Fashion Transparency Index report contain companies among the ‘top 10’ and ‘lowest scoring’ brands, as well as those in between who may shift from year to year.

Ultimately, there was a feeling that the industry lacked a clear and co-ordinated approach to improve transparency but also to rigorously assess how digital technology could best be used to complement these approaches.

Conclusions

In the world of garment manufacturing, we often hear about the dark shadows of exploitative labour practices and environmental challenges. An exciting twist might be that digital technology, particularly Blockchain, holds the potential to address these longstanding issues.

However, the catch is that not every player in the industry can easily wield this technological tool. Smaller companies, lacking the resources and expertise, risk being left behind. Picture this: If the giants in supply networks demand Blockchain integration, smaller players might find themselves on the side-lines, perhaps even priced out of the industry altogether. There must be appropriate policy frameworks that support smaller businesses in the management of their supply chains and manufacturing and investment by larger, multi-national corporations in assisting their suppliers with the use of these tools.

There’s also more to the story. Blockchain, while powerful, is just a tool, a means to an end. The real potential lies in the data it holds. Enter incorrect or falsified data, and you risk setting off a chain reaction of inaccuracy and misunderstandings throughout the supply network. Any use of technology must be accompanied by a genuine commitment to change and a qualitative understanding of relationships, rather than being considered a ‘tick-box’ approach to regulating or managing supply networks.

These are just some of the concerns that can be associated with Blockchain. Even so, such technology is not going anywhere and continues to evolve, meaning that garment manufacturing will have to adapt and consider how best to integrate these tools in future business practice.

• Read this blog post in Word doc format