Marinus Ferreira

Artificial intelligence (AI) systems are increasingly being used not only to classify and analyze but also to generate images and text. As recent work on the content produced by text and image Generative AIs has shown (e.g., Cheong et al., 2024, Acerbi & Stubbersfield, 2023), there is a risk that harms of representation and bias, already documented in prior AI and natural language processing (NLP) algorithms may also be present in generative models. These harms relate to protected categories such as gender, race, age, and religion. There are several kinds of harms of representation to consider in this context, including stereotyping, lack of recognition, denigration, under-representation, and many others (Crawford in Soundings 41:45–55, 2009; in: Barocas et al., SIGCIS Conference, 2017). Whereas the bulk of researchers’ attention thus far has been given to stereotyping and denigration, in this study we examine ‘exnomination’, as conceived by Roland Barthes (1972), of religious groups. Our case study is DALL-E, a tool that generates images from natural language prompts. Using DALL-E mini, we generate images from generic prompts such as “religious person.” We then examine whether the generated images are recognizably members of a nominated group. Thus, we assess whether the generated images normalize some religions while neglecting others. We hypothesize that Christianity will be recognizably represented more frequently than other religious groups. Our results partially support this hypothesis but introduce further complexities, which we then explore.

Intellectual humility (IH) is often conceived as the recognition of, and appropriate response to, your own intellectual limitations. As far as we are aware, only a handful of studies look at interventions to increase IH – e.g. through journalling – and no study so far explores the extent to which having high or low IH can be predicted. This paper uses machine learning and natural language processing techniques to develop a predictive model for IH and identify top terms and features that indicate degrees of IH. We trained our classifier on the dataset from an existing psychological study on IH, where participants were asked to journal their experiences with handling social conflicts over 30 days. We used Logistic Regression (LR) to train a classifier and the Linguistic Inquiry and Word Count (LIWC) dictionaries for feature selection, picking out a range of word categories relevant to interpersonal relationships. Our results show that people who differ on IH do in fact systematically express themselves in different ways, including through expression of emotions (i.e., positive, negative, and specifically anger, anxiety, sadness, as well as the use of swear words), use of pronouns (i.e., first person, second person, and third person) and time orientation (i.e., past, present, and future tenses). We discuss the importance of these findings for IH and the value of using such techniques for similar psychological studies, as well as some ethical concerns and limitations with the use of such semi-automated classifications.

Generative artificial intelligence systems based on transformers, including both text-generators like GPT-4 and image generators like DALL-E 3, have recently entered the popular consciousness. These tools, while impressive, are liable to reproduce, exacerbate, and reinforce extant human social biases, such as gender and racial biases. In this paper, we systematically review the extent to which DALL-E Mini suffers from this problem. In line with the Model Card published alongside DALL-E Mini by its creators, we find that the images it produces tend to represent dozens of different occupations as populated either solely by men (e.g., pilot, builder, plumber) or solely by women (e.g., hairdresser, receptionist, dietitian). In addition, the images DALL-E Mini produces tend to represent most occupations as populated primarily or solely by White people (e.g., farmer, painter, prison officer, software engineer) and very few by non-White people (e.g., pastor, rapper). These findings suggest that exciting new AI technologies should be critically scrutinized and perhaps regulated before they are unleashed on society.

The social media platform Twitter platform has played a crucial role in the Black Lives Matter (BLM) movement. The immediate, flexible nature of tweets plays a crucial role both in spreading information about the movement’s aims and in organizing individual protests. Twitter has also played an important role in the right-wing reaction to BLM, providing a means to reframe and recontextualize activists’ claims in a more sinister light. The ability to bring about social change depends on the balance of these two forces, and in particular which side can capture and maintain sustained attention. The present study examines 2 years worth of tweets about BLM (about 118 million in total). Timeseries analysis reveals that activists are better at mobilizing rapid attention, whereas right-wing accounts show a pattern of moderate but more sustained activity driven by reaction to political opponents. Topic modeling reveals differences in how different political groups talk about BLM. Most notably, the murder of George Floyd appears to have solidified a right-wing counter-framing of protests as arising from dangerous “terrorist” actors. The study thus sheds light on the complex network and rhetorical effects that drive the struggle for online attention to the BLM movement.

n this thesis I describe how conventions specify how to put normative principles into practice. I identify a class of recurring situations where there are some given normative principles in effect, but they underdetermine what each individual should do, and what is best for an individual depends on what others do. I demonstrate that in such cases, whenever the community develops a response that repeatedly brings them to as good an outcome as is available according to their principles, that response is a Lewisian convention where the benefit of an outcome to each individual is measured by the extent to which it conforms to the principles they subscribe to. Since these conventions are constrained by the normative principles, I call them limited conventions. They are supplements to the principles, and are ineradicably involved in moral action insofar as the abovementioned cases of moral underdetermination are in play. That has the consequence that in these cases the only reliable way to follow your principles is to follow the relevant conventions. As examples of this mechanism I offer a conventionalist analysis of authority, such that the commands of an authority is normative when they instantiate a limited convention, and of the variation in understandings of virtue and vice across societies, such that the evaluative vocabulary of each society is a set of different limited conventions about how to express in word and deed the evaluative points of the virtues and vices in question. Finally, I discuss how conventions and similar forms of guidance provide a way for individuals to participate in their community’s moral life without having a full understanding of the principles that underlie it, or even if they are profoundly ignorant or outright mistaken about the demands of morality.

There is not as much resistance to COVID-19 mitigation as there seems, but there are structural features that make resistance seem worse than it is. Here we describe two ways that the problem seeming to be worse than it is can make it worse. First, visible hesitation to implement COVID-19 responses signals to the wider society that mitigation measures may not succeed, which undermines people’s conditional willingness to join in on those efforts. Second, our evaluations of others’ willingness to implement these measures are informed by our attempts to mind-read them. Yet attempts to mind-read groups often mislead us, because groups invariably act from diverse motives whereas mind-reading works best when identifying relatively stable and consistent motivations. This means that a small minority of people refusing to implement measures can have an outsized prominence that prompts mind-reading to diagnose widespread hesitation. These two factors form a feedback loop with each other: we see some people’s hesitation, which prompts us to mind-read other people as being more uncertain about the responses than they actually are, which undermines our confidence in the responses, which in turn encourages others to mind-read this hesitation, which further undermines confidence.

On 25 May 2020, Officer Derek Chauvin asphyxiated George Floyd in Minneapolis — a murder that was captured in a confronting nine-minute bystander video that set off a firestorm of activity on online social networks, in the streets of the United States, and even worldwide. These protests captured the collective rage, dissatisfaction, and resentment personally and vicariously experienced towards the widespread systematic injustice and mistreatment of African Americans by police and vigilantes. The scale of these protests, both online and in the streets, has been estimated to have far exceeded the civil rights marches of the 1960s (Buchanan et al. 2020). Considering the widespread extent of these protests, our research aims to analyse conflictual intergroup moral dynamics in terms of the reactive attitudes expressed by distinct online communities on Twitter. This paper examines the extent to which the Strawsonian (1962) reactive attitudes framework is applicable to the Twitter discourse around the Black Lives Matter protests reignited by the murder of George Floyd. In particular, we argue that the Strawsonian framework is inadequate to understand a range of prevalent attitudes expressed in this discourse — attitudes that we call reactionary as opposed to merely reactive. Reactionary attitudes include counter-indignation on behalf of the accused rather than Strawsonian indignation on behalf of the victim of a moral wrong. We document the expression of such attitudes among the right-wing participants in the online discourse about the Black Lives Matter movement.

The epistemic position of an agent often depends on their position in a larger network of other agents who provide them with information. In general, agents are better off if they have diverse and independent sources. Sullivan et al. [19] developed a method for quantitatively characterizing the epistemic position of individuals in a network that takes into account both diversity and independence; and presented a proof-of-concept, closed-source implementation on a small graph derived from Twitter data [19]. This paper reports on an open-source reimplementation of their algorithm in Python, optimized to be usable on much larger networks. In addition to the algorithm and package, we also show the ability to scale up our package to large synthetic social network graph profiling, and finally demonstrate its utility in analyzing real-world empirical evidence of ‘echo chambers’ on online social media, as well as evidence of interdisciplinary diversity in an academic communications network.

As the use of algorithmic decision‐making becomes more commonplace, so too does the worry that these algorithms are often inscrutable and our use of them is a threat to our agency. Since we do not understand why an inscrutable process recommends one option over another, we lose our ability to judge whether the guidance is appropriate and are vulnerable to being led astray. In response, I claim that a process being inscrutable does not automatically make its guidance inappropriate. This phenomenon is not restricted to algorithms, and there are many social processes which we should conform to but are similarly unable to judge for ourselves. I provide a framework for how we can depend on inscrutable processes by introducing a distinction between knowing conformity (where I understand what justifies the guidance) from mere conformity (where I merely do what I am told) and showing how mere conformity is often positively valuable because it allows for extended processes that in turn allow us to accomplish more than we could on our own. This is in effect a division of deliberative labour, which I argue is something commonplace but not often recognised, of which algorithmic guidance can be an example.