Virtually Impossible: Obstacles to Generalizing between Simulated and Real Humans
Document Type
Article
Publication Title
PhilSci Archive
Publication Date
2-6-2025
Abstract/ Summary
The validity of a virtual human-based research methodology, in which simulated humans are used to generate knowledge about real humans, depends on substantiating multiple correspondence claims which are currently indefensible. One must substantiate that real and virtual humans are sufficiently similar with respect to their (1) control structures, (2) environments and embodied experiences, (3) adaptive histories and attunements, (4) social and cultural contexts, and (5) institutional contexts. If one’s confidence in any of these correspondences is undermined, then the foundation of this approach will crumble.
Unfortunately, technological limitations and our fragmentary understanding of minds will severely constrain the similarities between real and virtual humans for the foreseeable future. As a result, attempts to generalize empirical findings from virtual humans to real humans will prove ill-founded, and are likely to fail. Therefore, we believe that alternative research methodologies that focus on understanding mechanisms of mind more broadly, and cultivate the gradual acquisition of enabling technologies and engineering competences, are needed in the interim. We describe two such alternative approaches here, and speculate on their usefulness and viability in practice.