Privacy-Preserving Collective Intelligence: Federated Learning for Sensitive Domain Crowdsourcing

While Sun et al. (2024) and Edirimannage et al. (2024) have made advances in privacy protection for mobile crowdsourcing, their approaches focus on location and data privacy rather than the intellectual privacy needed for sensitive domains like healthcare or policy development. This research proposes a federated learning approach to collective intelligence that allows participants to contribute to knowledge aggregation without revealing their individual inputs or reasoning processes. Building on Policy Synth's AI-enhanced crowdsourcing (Bjarnason et al., 2024) but addressing its privacy limitations, the system would use homomorphic encryption and secure multi-party computation to enable meaningful aggregation of expert opinions while preserving confidentiality. This is particularly crucial for domains like Paiva's (2025) sustainable innovation platforms or Wu et al.'s (2022) emergency rumor control, where participants may have valuable insights but fear repercussions for sharing sensitive information. The innovation extends beyond typical federated learning by designing specialized aggregation protocols that work with qualitative, reasoning-based contributions rather than just numerical parameters. This could unlock collective intelligence applications in domains where privacy concerns have traditionally prevented effective crowdsourcing, potentially transforming how we address sensitive societal challenges through collective wisdom.

References:

1. Using Artificial Intelligence to Accelerate Collective Intelligence: Policy Synth and Smarter Crowdsourcing. R'obert Bjarnason, Dane Gambrell, Joshua Lanthier-Welch (2024). arXiv.org.
2. Harnessing Crowdsourced Innovation for Sustainable Impact: The Role of Digital Platforms in Mobilising Collective Intelligence. Teresa Paiva (2025). Platforms.
3. How Collective Intelligence Restrains the Rumor Spreading in Emergencies?. Pinmo Wu, Xiaowen Qu, Zhuoxi Zhong (2022). BCP Social Sciences & Humanities.
4. Bilateral Privacy Protection Scheme Based on Adaptive Location Generalization and Grouping Aggregation in Mobile Crowdsourcing. Xuelei Sun, Yingjie Wang, Peiyong Duan, Qasim Zia, Weilong Wang, Zhipeng Cai (2024). IEEE Internet of Things Journal.
5. QARMA-FL: Quality-Aware Robust Model Aggregation for Mobile Crowdsourcing. Shehan Edirimannage, Charitha Elvitigala, Ibrahim Khalil, Primal Wijesekera, X. Yi (2024). IEEE Internet of Things Journal.