Original title: Do People Follow AI Advice? Evidence from a Pension Portfolio Choice Experiment
Hongseok Choi, Jeongbin Kim, Matthew Kovach, Kyu-Min Lee +2 · August 2026
When artificial intelligence offers financial tips, do people actually listen? Researchers set out to see how AI guidance changes how people invest their retirement money. They tested this by having people build a test pension portfolio and then giving them an AI recommendation to see if they would change their picks. People could receive either a risky suggestion or a cautious one, sometimes paired with a brief explanation. Most participants did change their plans, and nearly all who changed moved closer to the computer advice. Even so, people only adopted about half of the suggested changes. Adding an explanation did not make people follow the advice any more closely. People listen to AI, but they still stick heavily to their own first choices.
Imagine a friend telling you how to season a stew. You taste it, add some salt, taste again, and add a little more—but you never add the full amount they suggested, because you trust your own first version too. Here, the friend's advice is the AI recommendation, and the stew is the way people divide up their retirement money: people move partway toward the suggestion but keep most of their original decision.
If AI tools increasingly recommend pension portfolios, the finding that people adopt only part of the advice—and that adding a rationale doesn't change this—matters for how much risk savers actually take. Plan sponsors and regulators who design defaults and robo-advice can't assume the advice is followed fully; the gap between recommendation and action shapes real retirement income.
The abstract doesn't say much about limitations, but it does describe the scope: a hypothetical pension balance, 400 employed adults in South Korean workplace defined contribution plans, and fixed AI recommendations. Results may not generalize to real money, other countries, or other types of advice. The abstract also does not measure long-term effects or whether participants' final portfolios were better or worse in any absolute sense.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
Do People Follow AI Advice? Evidence from a Pension Portfolio Choice Experiment — Hongseok Choi, Jeongbin Kim, Matthew Kovach et al.
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.