An arXiv preprint, 'How Far Can an Agent Push? Trust-Safety Margins in Live Language-Based Economic Games,' takes up a gap in standard game theory. In conventional models, a strategic action is either acceptable or not. The authors argue that in live, language-based economic games, the relevant question is also how long a counterpart will continue to accept the action.
To frame that question, they introduce the idea of a trust-safety margin: a measure of how far an agent can push before a live counterpart stops tolerating the behavior. The abstract is truncated after the opening, so the full methodology and results are not available from the source text.
Only this single source was provided, so there are no independent findings to compare. The paper's contribution, as stated, is to shift attention from whether an action is accepted to how long it remains accepted in live conversational economic settings.