Options
Can explainability metrics improve genetic programming? Lessons from 2048
Citation Link: https://doi.org/10.15480/882.18350
Publikationstyp
Conference Paper
Date Issued
2026-08-13
Sprache
English
TORE-DOI
Start Page
1135
End Page
1143
Citation
Genetic and Evolutionary Computation Conference, GECCO 2026
Contribution to Conference
Publisher DOI
Scopus ID
Publisher
Association for Computing Machinery (ACM)
ISBN of container
979-840072488-6
Genetic programming (GP) produces solutions that, while often less performant than neural networks, are uniquely amenable to analysis. In domains like game-playing, this explainability could reveal why certain policies succeed or fail-but such insights are rarely leveraged to improve the algorithms themselves. Here, we analyze a GP algorithm for 2048, testing whether structural, behavioral, or semantic metrics of evolved policies correlate with performance. While structural metrics showed no predictive power and behavioral features yielded ambiguous results, we identified a semantic feature that correlated with policy quality. Using this insight, we designed a mutation operator that improved performance. Though modest, this improvement suggests that explainability metrics can guide operator design, even when broader explanatory goals remain unmet. More broadly, our work highlights a potential advantage of GP: its solutions may be analyzable in ways that opaque methods like neural networks are not.
Subjects
behavioral analysis
explainability
game policies
genetic programming
structural analysis
DDC Class
006.3: Artificial Intelligence
Publication version
publishedVersion
Loading...
Name
3795101.3814717.pdf
Type
Main Article
Size
888.92 KB
Format
Adobe PDF