ABSTRACT. The implications of using machine learning (ML) tools that risk propagation of pernicious bias (that is, reflecting societal inequality) are of tremendous concern. The implications are even greater in healthcare where social determinants of health may independently contribute to healthcare inequalities. Although the mainstream perception appears to be that bias has arisen de novo and is attributable to ML per se, there is ample evidence to indicate that bias in ML models real-world patterns of social inequality. Given that ML-related techniques involve learning from associations within these extant, biased data, these applications require targeted attention to their ethical development and implementation to minimize the risk of unintended consequences stemming from propagation of bias. In this work, we briefly describe the range of ‘algorithmic fairness’ solutions offered within the fair ML field and how they operationalize and define ‘fairness.’ We explore how the efficacy of these solutions are likely highly limited in the field of healthcare ML by elucidating epistemic, empirical, and ethical considerations. Finally, we focus on how contributions from feminist critiques of science may inform a more ethically defensible path forward, and conclude with a set of recommendations for bias in healthcare ML.

ABSTRACT. There are many incompatible ways to measure fair outcomes for machine learning algorithms. The goal of this paper is to characterize rates of success and error across protected groups (race, gender, sexual orientation) as a distribution problem, and describe the possible solutions to this problem according to different normative principles from moral and political philosophy. These normative principles include: Consequentialism, Intent-Based and Compensation-Based Egalitarianism, Libertarianism, and Desert-Based Theories. Each principle will be applied to a sample risk-assessment classifier to demonstrate the philosophical arguments underlying different sets of fairness metrics.

ABSTRACT. In this paper, we analyze the relation between biased data-driven outcomes and practices of data annotation for vision models, by placing them in the context of the market economy. Understanding data annotation as a sense-making process, we investigate which goals are prioritized by decision-makers throughout the annotation of datasets. Following a qualitative design, the study is based on 24 interviews with relevant actors and extensive participatory observations, including several weeks of fieldwork at two companies dedicated to data annotation for machine learning in Buenos Aires, Argentina and Sofia, Bulgaria. The prevalence of market-oriented values over socially responsible approaches is argued based on three corporate priorities that inform work practices in this field: profit, standardization, and opacity. Finally, we introduce three elements, namely transparency, education, and regulations, aiming at developing ethics-oriented practices of data annotation, that could help prevent biased outcomes.

ABSTRACT. Concerns within the machine learning community and external pressures from regulators over the vulnerabilities of ma-chine learning algorithms have spurred on the fields of explainability, robustness, and fairness. Often, issues in explain-ability, robustness, and fairness are confined to their specific sub-fields and few tools exist for model developers to use to simultaneously build their modeling pipelines in a transparent, accountable, and fair way. This can lead to a bottleneck on the model developer’s side as they must juggle multiple methods to evaluate their algorithms. In this paper, we present a single framework for analyzing the robustness, fairness, and explainability of a classifier. The framework, which is based on the generation of counterfactual explanations through a custom genetic algorithm, is flexible, model-agnostic, and does not require access to model internals. The framework allows the user to calculate robustness and fairness scores for individual models and generate explanations for individual predictions which provide a means for actionable recourse(changes to an input to help get a desired outcome). This is the first time that a unified tool has been developed to address three key issues pertaining towards building a responsible artificial intelligence system.

ABSTRACT. What constitutes a `fair' electoral districting plan is a discussion dating back to the founding of the United States and, in light of several recent court cases, mathematical developments, and the approaching 2020 U.S. Census, is still a fiercely debated topic today. In light of the growing desire and ability to use algorithmic tools in drawing these districts, we discuss two prototypical formulations of fairness in this domain: drawing the districts by a neutral procedure or drawing them to intentionally induce a desirable electoral outcome. We then generate a large sample of districting plans for North Carolina and Pennsylvania and consider empirically how compactness and partisan symmetry, as instantiations of these frameworks, trade off with each other -- increasing the value of one of these necessarily decreases the value of the other.

ABSTRACT. Rideshare platforms, when assigning requests to drivers, tend to maximize profit for the system and/or minimize waiting time for riders. Such platforms can exacerbate biases that drivers may have over certain types of requests. We consider the case of peak hours when the demand for rides is more than the supply of drivers. Drivers are well aware of their advantage during the peak hours and can choose to be selective about which rides to accept. Moreover, if in such a scenario, the assignment of requests to drivers (by the platform) is made only to maximize profit and/or minimize wait time for riders, requests of a certain type (\eg from a non-popular pickup location, or to a non-popular drop-off location) might never be assigned to a driver. Such a system can be highly unfair to riders. However, increasing fairness might come at a cost of the overall profit made by the rideshare platform. To balance these conflicting goals, we present a flexible, non-adaptive algorithm, \lpalg, that allows the platform designer to control the profit and fairness of the system via parameters $\alpha$ and $\beta$ respectively. We model the matching problem as an online bipartite matching where the set of drivers is offline and requests arrive online. Upon the arrival of a request, we use \lpalg to assign it to a driver (the driver might then choose to accept or reject it) or reject the request. We formalize the measures of profit and fairness in our setting and show that by using \lpalg, the competitive ratios for profit and fairness measures would be no worse than $\alpha/e$ and $\beta/e$ respectively. Extensive experimental results on both real-world and synthetic datasets confirm the validity of our theoretical lower bounds. Additionally, they show that $\lpalg$ under some choice of $(\alpha, \beta)$ can beat two natural heuristics, Greedy and Uniform, on \emph{both} fairness and profit.

ABSTRACT. Public and private institutions must often allocate scare resources under uncertainty. Banks, for example, extend credit to loan applicants based in part on their estimated likelihood of repaying a loan. But when the quality of information differs across candidates (e.g., if some applicants lack traditional credit histories), common lending strategies can lead to undesirable disparities across groups. Here we consider a setting in which decision makers—before allocating resources—can choose to spend some of their limited budget further screening select individuals. We present a computationally efficient algorithm for deciding whom to screen that maximizes a standard measure of social welfare. Intuitively, decision makers should screen candidates on the margin, for whom the additional information could plausibly alter the allocation. We formalize this idea by showing the problem can be reduced to solving a series of linear programs. Both on synthetic and real-world datasets, this strategy improves utility, illustrating the value of targeted information acquisition in such decisions. Further, when there is social value for distributing resources to groups for whom we have a priori poor information—like those without credit scores—our approach can substantially improve the allocation of limited assets.

ABSTRACT. The ethical concept of fairness has recently been applied in machine learning (ML) settings to describe a wide range of constraints and objectives. When considering the relevance of ethical concepts to subset selection problems, the concepts of diversity and inclusion are additionally applicable in order to create outputs that account for social power and access differentials. We introduce metrics based on these concepts, which can be applied together, separately, and in tandem with additional fairness constraints. Results from human subject experiments support the proposed criteria. Social choice methods can additionally be leveraged to aggregate and choose preferable sets, and we detail how these may be applied.

ABSTRACT. Many arguments have concluded that our autonomous technologies must be intelligible, interpretable, or explainable, even if that property comes at a performance cost. In this paper, we consider the reasons why some property like these might be valuable, we conclude that there is not simply one kind of ‘intelligibility’, but rather different types for different individuals and uses. In particular, different interests and goals require different types of intelligibility (or explanations, or other related notion). We thus provide a typography of ‘intelligibility’ that distinguishes various notions, and draw methodological conclusions about how autonomous technologies should be designed and deployed in different ways, depending on whose intelligibility is required.

ABSTRACT. As machine learning black boxes are increasingly being deployed in domains such as healthcare and criminal justice, there is growing emphasis on building tools and techniques for explaining these black boxes in an interpretable manner. Such explanations are being leveraged by domain experts to diagnose systematic errors and underlying biases of black boxes. In this paper, we demonstrate that post hoc explanations techniques that rely on input perturbations, such as LIME and SHAP, are not reliable. Specifically, we propose a novel scaffolding technique that effectively hides the biases of a given classifier by allowing an adversarial entity to craft an arbitrary desired explanation. Our approach can be used to scaffold any biased classifier in such a way that its predictions on the input data distribution still remain biased, but the post hoc explanations of the scaffolded classifier look innocuous. Using extensive evaluation with multiple real-world datasets (including COMPAS), we demonstrate how extremely biased (racist) classifiers crafted by our framework can easily fool popular explanation techniques such as LIME and SHAP into generating innocuous explanations which do not reflect the underlying biases.

ABSTRACT. Bias in machine learning has manifested injustice in several areas, such as medicine, hiring, and criminal justice. In response, computer scientists have developed myriad definitions of fairness to correct this bias in fielded algorithms. While some definitions are based on established legal and ethical norms, others are largely mathematical. It is unclear whether the general public agrees with these fairness definitions, and perhaps more importantly, whether they understand these definitions. We take initial steps toward bridging this gap between ML researchers and the public, by addressing the question: does a non-technical audience understand a basic definition of ML fairness? We develop a metric to measure comprehension of one such definition--demographic parity. We validate this metric using online surveys, and study the relationship between comprehension and sentiment, demographics, and the application at hand.

ABSTRACT. Machine learning algorithms have gained widespread usage across a variety of domains, both in providing predictions to expert users and recommending decisions to everyday users. However, these AI systems are often black boxes, and end-users are rarely provided with an explanation. The critical need for explanation by AI systems has led to calls for algorithmic transparency, including the ``right to explanation" in the EU General Data Protection Regulation (GDPR). These initiatives presuppose that we know what constitutes a meaningful or good explanation, but there has actually been surprisingly little research on this question in the context of AI systems. In this paper, we (1) develop a generalizable framework grounded in philosophy, psychology, and interpretable machine learning to investigate and define characteristics of good explanation, and (2) conduct a large-scale lab experiment to measure the impact of different factors on people's perceptions of understanding, usage intention, and trust of AI systems. The framework and study together provide a concrete guide for managers on how to present algorithmic prediction rationales to end-users to foster trust and adoption, and elements of explanation and transparency to be considered by AI researchers and engineers in designing, developing, and deploying transparent or explainable algorithms.

ABSTRACT. As machine learning black boxes are increasingly being deployed in critical domains such as healthcare and criminal justice, there has been a growing emphasis on developing techniques for explaining these black boxes in a human interpretable manner. It has recently become apparent that a high-fidelity explanation of a black box ML model may not accurately reflect the biases in the black box. As a consequence, explanations have the potential to mislead human users into trusting a problematic black box. In this work, we rigorously explore the notion of misleading explanations and how they influence user trust in black-box models. More specifically, we propose a novel theoretical framework for understanding and generating misleading explanations, and carry out a user study with domain experts to demonstrate how these explanations can be used to mislead users. Our work is the first to empirically establish how user trust in black box models can be manipulated via misleading explanations.

ABSTRACT. Software is increasingly used to direct and manage critical aspects of all of our lives from how we get our news, to how we find a spouse, to how we navigate the streets of our cities. Beyond personal decisions, software plays a key role in regulated areas like housing, hiring and credit and major public areas like criminal justice and elections. Anyone who develops software knows how easy it is for there to be unintended defects. Bugs enter the systems at design time, during implementation and during deployment. Preventing, finding and fixing these flaws is a key focus of both industrial software development efforts as well as academic research in software engineering. In this paper, we discuss flaws in the larger socio-technical decision-making processes in which critical black-box software systems are approved, chosen, deployed and trusted. We use criminal justice software, specifically probabilistic genotyping (PG) software, as a concrete example. We describe how PG software systems, designed to do the same job, produce different results and discuss the impact of these differences on how the results are presented in court. We propose concrete changes to the socio-technical decision-making processes surrounding the use of PG software that could be used to incentivize debugging and improvements in the accuracy, fairness and reliability of these systems.

ABSTRACT. As AI Systems become increasingly autonomous they are expected to engage in complex moral decision-making processes. For the purpose of guidance of such processes theoretical and empirical solutions have been sought. In this research we integrate both theoretical and empirical lines of thought to address the matters of moral reasoning in AI Systems. We reconceptualize a metanormative framework for decision-making under moral uncertainty within the Discrete Choice Analysis domain and we operationalize it through a latent class choice model. The discrete choice analysis-based formulation of the metanormative framework is theory-rooted and practical as it captures moral uncertainty through a small set of latent classes. To illustrate our approach we conceptualize a society in which AI Systems are in charge of making policy choices. In the proof of concept two AI systems make policy choices on behalf of a society but while one of the systems uses a baseline moral certain model the other uses a moral uncertain model. It was observed that there are cases in which the AI Systems disagree about the policy to be chosen which we believe is an indication about the relevance of moral uncertainty.

ABSTRACT. Implicit in any rhetorical interaction—between humans or between humans and machines—are ethical codes, including a purpose code that provides the reason for our interacting in the first place. These ethical understandings are a key part of rhetorical context, the social situation in which communication happens but also the engine that drives communicative interaction. Such codes are usually invisible to AI writing systems because they do not typically exist in the databases the systems use to produce discourse. Can AI writing systems learn to learn rhetorical context, particularly the implicit codes for communication ethics? We see evidence that some systems do address issues of rhetorical context, at least in rudimentary ways. But we critique the information transfer communication model supporting many AI writing systems, arguing for a social context model that accounts for what is "not there" in the data set but that is critical for the production of meaningful, significant, and ethical communication. We offer two ethical principles to guide design of AI writing systems: transparency about machine presence and critical data awareness, a methodological reflexivity about omissions in the data that need to be provided by a human agent or accounted for in machine learning.

ABSTRACT. The recent unprecedented popularity of food recommender applications has raised several issues related to the ethical, societal and legal implications of relying on these applications. In this paper, in order to assess the relevant ethical issues, we rely on the emerging principles across the AI\&Ethics community and define them tailored context specifically. Considering the popular F-RS in the European market (YUKA, Fairtrade, Shopgun, etc..) cannot be regarded as personalised F-RS, we show how merely this lack of feature shifts the relevance of the focal ethical concerns. We identify the major challenges and propose a scheme for how explicit ethical agendas should be explained. We also argue how a multi-stakeholder approach is indispensable to ensure producing longterm benefits for all stakeholders. Given the argumentative nature of the paper, we limit ourselves to point to further research directions that could build on the defined ethical desiderata given by this paper from an AI architectural and, more importantly, from a legal perspective.

ABSTRACT. College admissions in the United States is carried out by a human-centered method of evaluation known as holistic review, which typically involves reading original narrative essays submitted by each applicant. The legitimacy and fairness of holistic review, which gives human readers significant discretion over determining each applicant's fitness for admission, has repeatedly been challenged in courtrooms and the public sphere. Using a unique corpus of 283,676 application essays submitted to a large, selective, state university system between 2015 and 2016, we assess the extent to which applicant demographic characteristics can be inferred from application essays. We find a relatively interpretable classifier (logistic regression) was able to predict gender and household income with high levels of accuracy. Findings suggest that auditing data might be useful in informing holistic review, and perhaps other evaluative systems, by checking potential bias in human or computational readings.

ABSTRACT. Even as public pressure mounts for technology companies to consider societal impacts of products, industries and governments in the AI race are demanding technical talent. To meet this demand, universities clamor to add technical artificial intelligence (AI) and machine learning (ML) courses into computing curriculum--but how are societal and ethical considerations part of this landscape? We explore two pathways for ethics content in AI education: (1) standalone AI ethics courses, and (2) integrating ethics into technical AI courses. For both pathways, we ask: What is being taught? As we train computer scientists who will build and deploy AI tools, how are we training them to consider the consequences of their work? In this exploratory work, we qualitatively analyzed 31 standalone AI ethics classes from 22 U.S. universities and 20 AI/ML technical courses from 12 U.S. universities to understand which ethics-related topics professors include in courses. We identify and categorize topics in AI ethics education, share notable practices, and note omissions. Our analysis will help AI educators identify what topics should be taught and create scaffolding for developing future AI ethics education.

ABSTRACT. A “rightful machine” is an explicitly moral, autonomous machine agent whose behavior conforms to principles of justice and the positive public law of a legitimate state. In this paper, I set out some basic elements of a deontic logic appropriate for capturing conflicting legal obligations for purposes of programming rightful machines. Justice demands that the prescriptive system of enforceable public laws be consistent, yet statutes or case holdings may often describe legal obligations that contradict; moreover, even fundamental constitutional rights may come into conflict. I argue that a deontic logic of the law should not try to work around such conflicts but, instead, identify and expose them so that the rights and duties that generate inconsistencies in public law can be explicitly qualified and the conflicts resolved. I then argue that a credulous, non-monotonic deontic logic can describe inconsistent legal obligations while meeting the normative demand for consistency in the prescriptive system of public law. I propose an implementation of this logic via a modified form of “answer set programming,” which I demonstrate with some simple examples.

ABSTRACT. Epistemology is the systematic philosophical examination of knowledge and is concerned with the nature of knowledge and how we acquire it (Lewis 1996). Amongst philosophers, there is consensus that for a mental state to count as a knowledge state it must minimally be a justified, true belief. If we presume, for the sake of this paper, that machine learning can be a source of knowledge, then it makes sense to wonder what kind of justification it involves.Prima facie, one might think that machine learning is epistemologically inscrutable (Selbst and Barocas 2018). After all, we don't usually have access to the black box in which models make decisions. Thus it might appear that machine learning decisions qua knowledge don’t have sufficient justification to count as knowledge. One might think this is because the models don't appear to have evidence or accessible reasons for their output. We suggest that this underlies the widespread interest in explainable or interpretable AI within the research community as well as the general public. Despite this inscrutability, machine learning is being deployed in human-consequential domains at a rapid pace. How do we rationalize on the one hand this seeming justificatory black box with the wide application of machine learning? We argue that, in general, people adopt implicit reliabilism regarding machine learning. Reliabilism is an epistemological theory of epistemic justification according to which a belief is warranted if it has been produced by a reliable processor method (Goldman 2012). In this paper, we explore what this means in the ML context. We then suggest that, in certain high-stakes domains with moral consequences, reliabilism does not provide another kind of necessary justification—moral justification.

ABSTRACT. There is more attention than ever before on the social implications of AI. In contrast to universalized paradigms of ethics and fairness, there is a move towards critical work that situates AI within the frame of social justice and human rights (“AI justice”). The geographical location of much of this critique in the West could however be engendering its own blind spots. AI’s global supply chain (data, labour, computation power, natural resources) today replicates geopolitical inequities, and the continued subordination of Global South countries. This paper draws attention to recent official policy narratives from India and United Nations Conference on Trade and Development (UNCTAD) aimed at influencing the role and place of these regions in the global political economy of AI. The flaws in these policies do not take away from the urgency of acknowledging colonial histories and the questions they raise of redistributive justice. Without a deliberate effort at initiating that conversation it is inevitable that mainstream discourse on AI justice will grow parallel to (and potentially undercut) demands emanating from Global South governments and communities.

ABSTRACT. The widespread deployment of machine learning models in high-stakes decision making scenarios requires a code of ethics for machine learning practitioners. We identify four of the primary components required for the ethical practice of machine learning: transparency, fairness, accountability, and reproducibility. We introduce Arbiter, a domain-specific programming language for machine learning practitioners that is designed for ethical machine learning. Arbiter provides a notation for recording how machine learning models will be trained, and we show how this notation can encourage the four described components of ethical machine learning.

ABSTRACT. Can effective international AI governance remain fragmented, or do we need a centralised international organisation for AI? We draw on the history of other international regimes to identify trade-offs in centralising AI governance. Some of these—(1) prevention of forum shopping; (2) policy coordination and political power; (3) efficiency and participation—speak for centralising governance. Others—(4) slowness and brittleness; (5) mutual supportiveness of decentralised approaches; (6) a lower buy-in threshold to—speak for decentralisation. Given these lessons, we conclude with two core recommendations. First, the outcome will depend on the details. A well-designed centralised regime would be optimal. But locking-in an inadequate structure one may be a fate worse than fragmentation. Second, for now fragmentation will likely persist. This should be closely monitored to see if it is self-organising or simply inadequate.

ABSTRACT. The ‘robot rights’ debate, and its related question of ‘robot responsibility’, invokes some of the most polarized positions in AI ethics. While some advo-cate for granting robots rights on a par with human beings, others, in a stark opposition argue that ro-bots are not deserving of rights but are objects that should be our slaves. Grounded in post-Cartesian philosophical foundations, we argue not just to de-ny robots ‘rights’, but to deny that robots, as arti-facts emerging out of and mediating human being, are the kinds of things that could be granted rights in the first place. Once we see robots as mediators of human being, we can understand how the ‘ro-bots rights’ debate is focused on first world prob-lems, at the expense of urgent ethical concerns, such as machine bias and machine elicited human labour exploitation, both impacting society’s least privileged individuals. We conclude that, if human being is our starting point and human welfare is the primary concern, the negative impacts emerging from machinic systems, as well the lack of taking responsibility by people designing, buying and de-ploying such machines, remains the only relevant ethical discussion in AI.

ABSTRACT. Autonomous vehicles (AVs) and accidents they are involved in attest to the urgent need to consider the ethics of AI. The question dominating the discussion has been whether we want AVs to behave in a ‘selfish’ or utilitarian manner. Rather than considering modeling self-driving cars on a single moral system like utilitarianism, one possible way to approach programming for AI would be to reflect recent work in neuroethics. The Agent-Deed-Consequence (ADC) model provides a promising account while also lending itself well to implementation in AI. The ADC model explains moral judgments by breaking them down into positive or negative intuitive evaluations of the Agent, Deed, and Consequence in any given situation. These intuitive evaluations combine to produce a judgment of moral acceptability. This explains the considerable flexibility and stability of human moral judgment that has yet to be replicated in AI. This paper examines the advantages and disadvantages of implementing the ADC model and how the model could inform future work on ethics of AI in general.

ABSTRACT. Today the societal influence of Artificial Intelligence (AI) is significantly widespread and continues to raise novel human rights concerns. As ‘control’ is increasingly ceded to AI systems, and potentially Artificial General Intelligence (AGI) humanity may be facing an identity crisis sooner rather than later, whereby the notion of ‘intelligence’ no longer remains solely our own. This paper characterizes the problem in terms of an emerging responsibility gap and loss of control and proposes a relational shift in our attitude towards AI. This shift can potentially be achieved through value alignment by incorporating Indigenous perspectives into AI development. The value of Indigenous perspectives has not been canvassed widely in the literature and becomes clear when considering the existence of well-developed epistemologies adept at accounting for the non-human, a task that defies Western anthropocentrism. Accommodating the non-human AI by considering it as part of our network is a step towards building a symbiotic relationship with AI. It is argued that in order to co-exist, as AGI potentially questions our fundamental notions of what it means to have human rights, we find assistance in well tested Indigenous traditions such as the Hawaiian (kānaka maoli) and Lakota ontologies.

ABSTRACT. As the transformative potential of AI has become increasingly salient as a matter of public and political interest, there has been growing discussion about the need to ensure that AI broadly benefits humanity. This in turn has spurred debate on the social responsibilities of large technology companies to serve the interests of society at large. In response, ethical principles and codes of conduct have been proposed to meet the escalating demand for this responsibility to be taken seriously. As yet, however, few institutional innovations have been suggested to translate this responsibility into legal commitments which apply to companies positioned to reap large financial gains from the development and use of AI. This pa- per offers one potentially attractive tool for addressing such issues: the Windfall Clause, which is an ex ante commitment by AI firms to donate a significant amount of any eventual extremely large profits. By this we mean an early commitment that profits that a firm could not earn without achieving fundamental, economically transformative breakthroughs in AI capabilities will be donated to benefit humanity broadly, with particular attention towards mitigating any downsides from deployment of windfall-generating AI.

ABSTRACT. In federated learning, the federation crowdsources data owners to share their local data by leveraging privacy preserving technologies in order to build a federated model. The model can achieve better performance than that of training with just the local data. However, in FL, participants need to incur some cost for contributing to the FL models with their local datasets. The training and commercialization of the models will take time. Thus, there will be some delays before the federation has enough budget to pay back the participants. This temporary mismatch between contributions and rewards has not been accounted for by existing payoff-sharing schemes. To address this limitation, we propose the Federated Learning Incentivizer (FLI) payoff-sharing scheme in this paper. The scheme dynamically divides a given budget in a context-aware manner among data owners in a federation by jointly maximizing the collective utility while minimizing the inequality among the data owners, in terms of the payoff gained by them and the waiting time for receiving payoffs. Extensive experimental comparisons with five state-of-the-art payoff-sharing schemes show that FLI is the most attractive to high quality data owners and achieves the highest expected revenue for a data federation.

ABSTRACT. Jakarta’s roads are green, filled as they are with the fluorescent green jackets, bright green logos and fluttering green banners of basecamps created by the city’s digitized, ‘online’ motorbike-taxi drivers (ojol). These spaces function as waiting posts, regulatory institutions, information networks and spaces of solidarity for the ojol working for mobility-app companies, Grab and GoJek. Their existence though, presents a puzzle. In the world of on-demand matching, literature predicts an isolated, atomized, disempowered digital worker. Yet, ojol basecamps persist both in the physical world and digital realm, complete with quirky names, emblems, social media accounts and even their own emergency response service. In fact, their solidarity actively revolves around the identity of using a smartphone to generate employment, and a key part of the imagined networks of community occur online through WhatsApp groups. This paper asks, under what conditions are digital workers able to create collective structures of solidarity, even as app-mediated work may force them towards an individualized labor regime? Are Jakarta’s digital labor collectives a reflection of its social context, a product of technological change, or a result of interactions between both? Increasingly, academic world has started paying attention to forms of digital labor organization. The aim of this project is to empirically tease out the bi-directional conversation between globalizing digital platforms and social norms, civic culture and labor market conditions in non-western urban spaces, which allow for particular forms of digital worker resistances to emerge. It recovers power for the worker, who provides us with a path to resisting algorithmization of work while still participating in it.

ABSTRACT. The recent wave of AI and automation has been argued to differ from previous General Purpose Technologies (GPTs), in that it may lead to rapid change in occupations' underlying task requirements and persistent technological unemployment. In this paper, we apply a novel methodology of dynamic task shares to a large dataset of online job postings to explore how exactly occupational task demands have changed over the past decade of AI innovation, especially across high, mid and low wage occupations. Notably, big data and AI have risen significantly among high wage occupations since 2012 and 2016, respectively. We built an ARIMA model to predict future occupational task demands and showcase several relevant examples in Healthcare, Administration, and IT. Such task demands predictions across occupations will play a pivotal role in retraining the workforce of the future.

ABSTRACT. In this paper we present a setting for examining the relation between the distribution of research intensity in AI research and the relevance for a range of work tasks (and occupations) in current and simulated scenarios. We perform a mapping between labour and AI using a set of cognitive abilities as an intermediate layer. This setting favours a two-way interpretation to analyse (1) what impact current or simulated AI research activity has or would have on labour-related tasks and occupations, and (2) what areas of AI research activity would be responsible for a desired or undesired effect on specific labour tasks and occupations. Concretely, in our analysis we map 59 generic labour-related tasks from several worker surveys and databases to 14 cognitive abilities from the cognitive science literature, and these to a comprehensive list of 328 AI benchmarks used to evaluate progress in AI techniques. We provide this model and its implementation as a tool for simulations. We also show the effectiveness of our setting with some illustrative examples.