Identifying Proposers’ Behavioral Patterns in Human-AI Economic Interactions
Bibliographic record
Abstract
Understanding how humans adapt their decision-making in economic interactions with artificial intelligence (AI) is essential for building socially attuned AI agents. In this study, we analyzed human proposers’ behavior in the Ultimatum Game (UG) using interpretable behavioral features and supervised machine learning models to classify strategic proposer types (Fair, Selfish, Learner, Tit-for-Tat). Using data from human–human and human–AI interactions in a UG experiment, we uncover context-sensitive patterns in proposer behavior. The analyses revealed that machine learning models—especially Random Forest (RF) and Neural Network (NN) can reliably identify behavioral strategy types with high accuracy across both interaction contexts. Classification was slightly more stable in the human condition, though the strongest models also generalized well to AI interactions. In contrast, simpler models such as Logistic Regression (LR) and Support Vector Machine (SVM) showed reduced performance in the AI condition, suggesting greater variability in human behavior when interacting with artificial agents. These findings suggest that while strategic behavior remains recognizable, collaboration with AI partners introduces greater variability, potentially due to expectancy violations or ambiguous fairness norms. Outcomes in human–AI interactions appear to depend on whether the context is cooperative (e.g., fair or tit-for-tat strategies) or competitive (e.g., exploitative or self-maximizing behavior). These insights can inform AI design, particularly when it comes to developing systems that interact more equitably and ethically with humans.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.009 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".