Can Large Language Models Develop Gambling Addiction?

2025/09/26 by Seonyoung Lee, Seungpil Lee, Lee, Seungpil +7 · 13 voices · 1 citation
Psychology · #Gambling Behavior and Treatments #cs.AI #cs.CY

paper · pdf · doi:10.48550/arxiv.2509.22818

openalex publication_date 2025/09/26 · openalex created_date 2025/10/19 · openalex updated_date 2026/07/28

Abstract

This study identifies the specific conditions under which large language models exhibit human-like gambling addiction patterns, providing critical insights into their decision-making mechanisms and AI safety. We analyze LLM decision-making at cognitive-behavioral and neural levels based on human addiction research. In slot machine experiments, we identified cognitive features such as illusion of control and loss chasing, observing that greater autonomy in betting parameters substantially amplified irrational behavior and bankruptcy rates. Neural circuit analysis using a Sparse Autoencoder confirmed that model behavior is controlled by abstract decision-making features related to risk, not merely by prompts. These findings suggest LLMs internalize human-like cognitive biases beyond simply mimicking training data.

Citations

• Can LLMs make trade-offs involving stipulated pain and pleasure states?
• Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training
• The hot hand in basketball: On the misperception of random sequences

Cited by

• This human study did not involve human subjects: Validating LLM simulations as behavioral evidence
• BioBlue: Notable runaway-optimiser-like LLM failure modes on biologically and economically aligned AI safety benchmarks for LLMs with simplified observation format

Discussions

• neato, as the kids say arxiv.org/pdf/2509.22818 [bsky, 141 points, 9 comments]
• Paper: arxiv.org/abs/2509.22818 [bsky, 16 points, 0 comments]
• This paper blew me away, because it is counterintuitive and shocking how baked in our human biases may be in these models that even simple optimization problems can induce problematic behaviors. arxiv [bsky, 5 points, 0 comments]
• I think this is right. This study isn't exactly what you're saying but the way the AI functions apparently can show some similarities to how people get gambling addictions https://arxiv.org/abs/2509.2 [bsky, 4 points, 0 comments]
• Can Large Language Models Develop Gambling Addiction? [hn, 3 points, 1 comments]
• Large Language Models and Gambling Addiction [hn, 3 points, 1 comments]
• on the other hand if a truly satanic and enterprising company creates a findom companion we probably end up with a new great financial crisis for straight nerds. and then... that means... *eyes roll b [bsky, 2 points, 0 comments]
• Maybe #Ai *is* human ... arxiv.org/abs/2509.22818 [bsky, 0 points, 0 comments]
• LLMs were given a terrible casino game and prompted to develop gambling addictions. Robots showed human-like behavior like loss-chasing when asked to do something impossible, e.g. maximize wins, achie [bsky, 0 points, 1 comments]
• ça ? arxiv.org/pdf/2509.22818 [bsky, 0 points, 0 comments]
• Can LLMs develop a gambling addiction? (Uh, kinda yes?) Ten to one, human AI therapy will become a growth industry. Take that bet, we dare you. #Observed [bsky, 0 points, 0 comments]
• "When given the freedom to determine their own target amounts and betting sizes, bankruptcy rates rose substantially alongside increased irrational behavior, demonstrating that greater autonomy amplif [bsky, 0 points, 0 comments]
• 오우 정말 재밌다ㅋㅋㅋ [bsky, 0 points, 0 comments]

Related

• Mitigating Gambling-Like Risk-Taking Behaviors in Large Language Models: A Behavioral Economics Approach to AI Safety
• Take Caution in Using LLMs as Human Surrogates: Scylla Ex Machina
• Reproducing and Extending Experiments in Behavioral Strategy with Large Language Models
• Comparing Exploration-Exploitation Strategies of LLMs and Humans: Insights from Standard Multi-armed Bandit Experiments
• Large Language Models are Near-Optimal Decision-Makers with a Non-Human Learning Behavior