# The relationship between trust in AI and trustworthy machine learning technologies

**Authors:** Ehsan Toreini, Mhairi Aitken, Kovila Coopamootoo, Karen Elliott, Carlos Gonzalez Zelaya, Aad van Moorsel  
**arXiv:** [1912.00782](https://arxiv.org/abs/1912.00782)  
**Date:** 2019-11-27  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/1912.00782v2/

## Key points

- Relationship between trust in AI and trustworthy machine learning technologies explored
- Importance of understanding how machine learning technologies impact trust emphasized
- Systematic approach proposed to relate social science concepts of trust with AI technologies
- Trust conceptualized using ABI framework and mapped onto different technologies
- Four categories of machine learning technologies focused on: Fairness, Explainability, Auditability, and Safety (FEAS)
- Examination of trust throughout the life cycle of AI-based systems
- Introduction of "Chain of Trust" concept to discuss technological needs for trust at different stages
- Connection between FEAS and international Principled AI policy and technology frameworks established
- Comprehensive analysis provided on how machine learning technologies impact trust in AI-based systems
- Insights offered on building trustworthy systems by aligning social science concepts of trust with technological qualities.

## Summary

This paper explores the relationship between trust in AI and trustworthy machine learning technologies. The authors emphasize the importance of understanding how machine learning technologies impact trust in order to build AI-based systems that users and the public can trust. To guide technology developments, they propose a systematic approach that relates social science concepts of trust with the technologies used in AI-based services and products. The authors conceptualize trust using the ABI (Ability, Benevolence, Integrity) framework and map these qualities onto different technologies. They specifically focus on four categories of machine learning technologies: Fairness, Explainability, Auditability, and Safety (FEAS). The paper discusses whether these technologies possess the required qualities for building trust. The concept of trust is examined throughout the life cycle of AI-based systems. The authors introduce the concept of a "Chain of Trust" to discuss technological needs for trust at different stages of the life cycle. This highlights the importance of considering trust not only during system development but also during deployment and maintenance. In addition, the authors relate FEAS to various international Principled AI policy and technology frameworks that have emerged in recent years. By doing so, they establish connections between their proposed approach and existing frameworks, further validating its relevance and applicability. Overall, this paper provides a comprehensive analysis of how machine learning technologies impact trust in AI-based systems. It offers valuable insights into building trustworthy systems by considering social science concepts of trust and aligning them with specific technological qualities.