Abstract: This paper examines the role of moral reasoning in shaping the adoption of artificial intelligence (AI) in the labor market. We focus on a phenomenon we term "moral resistance" to AI, which we measure across all US occupations, assessing the general population's moral resistance to the use of AI in various occupations.

As AI advances, its capacity to automate and enhance tasks expands, profoundly transforming the labor market. While previous research has concentrated on AI's technical capabilities for various tasks, there's a notable chasm between what AI can do and society's willingness to embrace it. Even when AI proves competent, there can be vigorous opposition to its use.

To examine the social processes shaping the adoption of AI, we focus on a stubborn type of opposition we call "moral resistance." Drawing on the common distinction in moral psychology between consequentialist and deontological reasoning, we propose that deontological objections to AI—those rooted in moral principles rather than outcomes—signal a more enduring resistance to AI, one that is likely to remain steadfast despite advancements in AI that promise better accuracy, explainability, and a reduction in bias.

Initial findings from an online survey covering 1,016 occupations reveal significant heterogeneity in moral resistance to AI across occupations. We also find only a weak correlation between AI's technical aptness for an occupation and the level of moral resistance it encounters. These results highlight "off-diagonal" occupations that, despite their technical compatibility with AI, face moral objections. We also report early results from the next phase of our study, which is to develop and validate a scale for measuring moral resistance to AI.

Our research shines a light on the potential paths and obstacles for integrating AI into the workforce, highlighting the crucial role of moral cognition and reasoning in shaping the future of work. More broadly, our study underscores the importance of social, legal, and political processes in determining how AI is woven into the fabric of the job market.

Digital platforms exert significant control over producers through algorithms, yet how producers strategically respond remains underexplored. Through an 18-month ethnographic study of content creators on Twitch, we uncover a novel practice of "interhoming" - the strategic integration of multiple platforms' algorithms to achieve producers' unique goals. Unlike multihoming, where producers use platforms as substitutes, interhoming leverages platforms as complements. This practice exemplifies "algorithmic entrepreneurship," where producers creatively combine algorithms across competing platforms to create and capture value. Our findings challenge the view of producers as passive recipients of algorithmic control, instead revealing them as active agents who strategically navigate a complex algorithmic landscape. By shifting focus from value capture to value creation, we provide a more holistic understanding of how algorithms shape economic opportunity. This study contributes to literature on algorithmic control and digital entrepreneurship by illuminating how producers exercise agency within multi-algorithmic environments, crafting sophisticated combinations where algorithms act as complements or substitutes.

Abstract: Why are some combinations of categories more viable than others? Prior work has examined conditions under which people are generally more or less tolerant of offerings that span multiple categories, but this leaves open the question of why some combinations of categories remain more viable than others. I develop and test a theory in which people evaluate category-spanning offerings using a standard of theoretical coherence—whether a particular combination coheres with their theory of value. To measure coherence from the perspective of different theories of value, I develop a measure using word embedding models that can readily be applied to other settings. Key implications of the theory are validated using data on the video game repertoires of livestreamers on the platform Twitch.tv. By providing an alternative explanation of how categories constrain strategic choice and valuation, the standard of theoretical coherence allows us to move beyond debates about typicality and specialist vs. generalist premiums to focus on which particular combinations of categories are viable.

Abstract: Entrepreneurial experiments are thought to improve outcomes because they avoid commitment and create real options, allowing the entrepreneur to identify new, attractive markets as they refine their idea. This paper documents a case where audiences demand prior commitment before they will tolerate experimentation. Building on this case, I develop theory that identifies a tradeoff faced by entrepreneurs who wish to use experiments to test the viability of their ideas. The tradeoff relates to the need to commit to an audience before the audience will tolerate experimentation. On one hand, entrepreneurs who enact a strong commitment to their audience find that their audience is more tolerant of experimentation, which in turn allows them to trial cruder prototypes and a run more experiments than a weakly committed entrepreneur. On the other hand, this commitment entails an obligation to heed feedback, resulting in a trajectory that reinforces appeal to the current audience at the expense of growth in new markets. I validate two key implications of this theory using data from a large livestreaming platform. I find that streamers who enact a strong commitment strategy experience less viewer drop-off when experimenting with new additions to their repertoire but that each addition leads to less growth in viewership than streamers who enact a weak commitment strategy. This implies that the entrepreneurial experimentation must, at least in some contexts, be understood as a relational process and that the commitment required for audiences to tolerate experimentation results in a tradeoff between growth on the intensive vs. extensive margin.

Abstract: Since Gouldner (1960), social scientists have seen the generalized norm of reciprocity (NOR) as an internalized norm that provides a key “starting mechanism” for exchange among strangers. However, we observe that the NOR is more frequently invoked among strong ties—i.e., where it is presumably extraneous and even off-putting. To explain this difference and provide a stronger foundation for understanding how the NOR facilitates social exchange, we rework Leifer’s (1988) theory of “local action” so that it encompasses a broader array of strategic motives. Whereas Leifer explained why strangers will avoid invoking the NOR in a bid to limit risky claims to status, our theory entails that such avoidance can be expected whenever instrumental goals are salient and mutual commitment is low. And since the costs of imbalance are high precisely when mutual commitment is high (as the parties may be “stuck” with the imbalance “forever”), parties to strong ties should be more likely to invoke the NOR. Two online vignette experiments provide preliminary support for the theory.