Human Heuristics for Autonomous Agents

TL;DR

This paper studies how autonomous agents use heuristics to decide when to verify information with a central database, balancing trust, cost, and risk of infection in a system with potential message corruption.

Contribution

It introduces a model of autonomous agents employing heuristics for trust and verification, considering costs and information decay, to analyze information integrity in corrupted message environments.

Findings

Agents balance verification costs and trust to prevent infection.

Information decay influences agents' verification strategies.

System behavior varies with fixed or spontaneously appearing corrupted messages.

Abstract

We investigate the problem of autonomous agents processing pieces of information that may be corrupted (tainted). Agents have the option of contacting a central database for a reliable check of the status of the message, but this procedure is costly and therefore should be used with parsimony. Agents have to evaluate the risk of being infected, and decide if and when communicating partners are affordable. Trustability is implemented as a personal (one-to-one) record of past contacts among agents, and as a mean-field monitoring of the level of message corruption. Moreover, this information is slowly forgotten in time, so that at the end everybody is checked against the database. We explore the behavior of a homogeneous system in the case of a fixed pool of spreaders of corrupted messages, and in the case of spontaneous appearance of corrupted messages.