Introduction
In negotiation, information exchange is a double-edged sword. While sharing preferences helps parties find "integrative" (win-win) solutions, it also creates an opening for malicious actors. A deceptive negotiator can use a "bogey tactic"—feigning high interest in an item they do not actually value—to "concede" that item later in exchange for something they truly want. This paper validates a game-theoretic model of "strategic misrepresentation," showing how deceptive negotiators exploit the human desire for fairness to secure lopsided deals.
The Misrepresentation Game
Previous research introduced the misrepresentation game, a mathematical framework for this behavior. The model treats deception as an optimization problem: a malicious agent aims to claim a false set of preferences such that the resulting "fair" deal (the Nash bargaining solution) actually maximizes their true utility. This paper seeks to bridge the gap between theory and reality by testing whether this model holds up in unstructured, face-to-face human dialogue.
Experimental Methodology
The study utilized the "Auction Wars" corpus, consisting of 113 dyads engaged in simulated negotiations over storage locker items (records, lamps, and a painting). The researchers compared two distinct scenarios: distributive tasks, where interests are directly opposed, and integrative tasks, where potential for mutual gain exists.
To detect deception, the researchers employed two automated linguistic heuristics:
- Issue-Sentiment (Words): An algorithm counts positive and negative statements about specific items to infer what a negotiator claims to value.
- Issue-Ratio (Deeds): An algorithm analyzes actual offers to infer what a negotiator actually values, assuming people generally try to keep more of what they find important.
Key Insights: The "Fixed-Pie Lie"
In win-win scenarios, the most successful liars are those who pretend their interests are identical to their opponent's. By claiming to want exactly what the other person wants, the liar makes their eventual "concessions" look incredibly generous and fair. Human negotiators have a strong "fixed-pie bias"—they often assume the other person wants what they want. Malicious actors exploit this cognitive shortcut to appear honest while being extractive.
Interestingly, in win-win scenarios, liars were actually rated as more honest than truthful negotiators. Because their lies aligned with the opponent's expectations, the deception remained undetected by human partners.
Results and Detection Accuracy
The research confirmed that lying is highly profitable in integrative negotiations, with liars earning significantly more than non-liars (Mean = 63.33 vs. 55.12 lottery tickets). Furthermore, 88% of the lies told in these tasks were "fixed-pie lies," exactly as the misrepresentation game predicted.
While humans failed to detect these lies, the researchers' "rank distance" metric—the divergence between a negotiator's words and their deeds—successfully identified deceptive behavior. As shown in the data, liars exhibited a significantly higher divergence than honest participants.
Conclusion and Future Work
This study demonstrates that deception creates a measurable behavioral signature. Liars must eventually make offers that reflect their true interests to win, creating a "deeds" profile that contradicts their "words."
Future work will address current limitations, such as the assumption of issue independence and the narrow focus on preference rankings. Additionally, researchers aim to design "defensive" AI agents capable of identifying malicious actors and steering negotiations back toward truly fair outcomes.
Summary updated