SearcharxivSearch

arXiv subjects

Leon Houf

Publications and source records attributed to Leon Houf.

3 recordsLinked to original sources

Going Beyond the In-/Out-Group Dichotomy: Investigating Altruism towards Middle-Groups

In-group favouritism and out-group hostility are well-documented, but real-world group settings rarely fit a simple dichotomy. Often, a "middle-group" shares some identity markers with the in-group without fully belonging to it. How do people treat such intermediate groups? We address this question using a formal identity marker framework and a multi-lab online experiment (N = 376) with a private allocation task immune to reputation effects and demand characteristics. We find that a middle-group can be treated neutrally, i.e., distinct from both in-group favouritism and out-group hostility, but only under specific conditions. When the middle-group shares two identity markers with the in-group, including university affiliation, a complete hierarchy emerges: the in-group is favoured, the middle-group is treated according to the allocation rule, and the out-group bears the loss. However, when the middle-group shares only one marker, it is treated indistinguishably from the out-group, regardless of which marker is shared. Furthermore, differentiated behaviour evolves only after participants have re-encountered the experimental instructions and group constellation in a second phase, not gradually across rounds. These findings demonstrate that participants can perceive beyond binary in-/out-group categorisation, but such perception requires both sufficient identity overlap and repeated exposure to the group setting. Our identity marker framework provides a tool for systematically studying group relations in more complex settings.

econ.GN

Access to Live AI Advice and Behavior Under Risk: An Incentivized Experiment

Generative AI has become an everyday advisor, and the systems people consult are live and interactive, not pre-scripted. We ask whether access to such a system changes behavior under risk. In an incentivized experiment (N = 158), participants made lottery choices with an optional decision aid presented as a conventional pre-written tool, a live one-shot AI, or a live interactive AI they could query, with information format held equivalent across conditions. Risk preferences are elicited via DOSE. We find no evidence that access to a live AI advisor changes risk aversion.

econ.GN

The Internal Anatomy of Strategic Choice in Large Language Models

Large language models act as strategic agents and models of human choice, yet choosing like a strategic agent does not mean computing like one. We recorded activations from four open-weight models --- dense and mixture-of-experts, including a matched base--instruct pair --- in one-shot play of 144 strict ordinal $2\times2$ games. We followed a prespecified incentive from prompt, through activations, to choice. Dense models mirrored the unadjusted human decline with game complexity. Incentive and choice were detectable in every model, but models differed in whether incentive reached the choice, aligned with it and, where tested, whether strengthening it shifted preference. The base and instruction-tuned Qwen2.5 models chose almost identically at baseline yet differed in whether incentive reached choice. Fixed decision cues were distinguishable internally but changed choices selectively. Similar behaviour can rest on different computation; post-training can reshape the path from represented incentive to decision while leaving behaviour and decodable information largely intact.

cs.AI