Crowdsourcing Impacts: Exploring the Utility of Crowds for Anticipating Societal Impacts of Algorithmic Decision Making

Crowdsourcing can effectively identify diverse societal impacts of algorithms: The study demonstrates that involving crowds in participatory foresight successfully uncovers multiple types of algorithmic impacts across four categories—perceived valence, societal domains, specific impact types, and ethical concerns—suggesting this approach captures a broader range of issues than expert-only assessments might.

Cognitive diversity of crowds reveals interconnected impacts: The research shows that crowdsourcing illuminates patterns and connections between different types of impacts, indicating that diverse perspectives help identify how algorithmic harms may interact and compound across different areas of society.

This method complements existing algorithmic assessment approaches: Rather than replacing current evaluation methods, crowdsourcing for impact anticipation works alongside other assessment techniques, offering a participatory dimension that can catch societal concerns that traditional technical or expert-led reviews might miss.

Participatory foresight should be integrated earlier in algorithm development: The paper suggests that involving broader publics in anticipating impacts at different stages of algorithm development—particularly governmental decision-making tools—can help researchers and designers proactively consider ethical implications before deployment.

Government algorithms warrant particular scrutiny through public input: By focusing on governmental algorithmic decision-making tools, the research highlights that algorithms affecting public services and policy decisions have especially important societal implications that benefit from diverse public perspectives during the design phase.