Original title: Access to Live AI Advice and Behavior Under Risk: An Incentivized Experiment
Paul Althaus, Leon Houf, Christiane Schwieren · September 2026
People now regularly ask generative artificial intelligence for advice. These AI systems answer in real time rather than using fixed scripts, but we do not know if getting guidance from them changes how much risk people take. To test this, researchers had volunteers make real choices between risky lotteries. Participants could use an optional helper. One group saw a conventional pre-written tool. Another group saw a live one-shot AI. A third group saw a live interactive AI they could actively chat with and question. All tools shared the same information format. The researchers found no evidence that having access to a live AI advisor alters people's risk aversion. Getting dynamic advice does not appear to change how people handle risk.
You are at a fair deciding whether to play a risky game. You can ask a friend for advice. The friend might give a one-time answer or let you ask questions. The researchers found that whether the friend is interactive or not doesn't change how willing you are to play the game. In this paper, the game is a lottery choice, and the friend is an AI advisor.
As people increasingly rely on AI for decisions involving risk, understanding whether interactive advice changes their behavior is crucial. This study suggests that live AI may not shift risk preferences, which could inform how we design AI advisors for financial or health decisions. It also raises questions about the effectiveness of AI in influencing human risk-taking.
The abstract doesn't say, but the study had 158 participants, used a specific risk elicitation method (DOSE), and a particular AI setup; results may not generalize to other contexts or AI systems.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
Access to Live AI Advice and Behavior Under Risk: An Incentivized Experiment — Paul Althaus, Leon Houf, Christiane Schwieren
Abstract licence: CC0
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.