Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

An Empirical Exploration of Trust Dynamics in LLM Supply Chains

Agathe Balayn, Mireia Yurrita, Fanny Rancourt, Fabio Casati, Ujwal Gadiraju

TL;DR

This paper broadens the study of trust in AI by examining trust dynamics across the entire LLM supply chain rather than limiting analysis to the user–LLM dyad. Through in-situ qualitative interviews with 71 practitioners across 10 organizations, it reveals a diverse set of trustors and trustees—including humans and various technical artifacts—and documents how trust cues based on the ABI framework (ability, benevolence, integrity) vary across actors. The findings show that trust relationships span downstream and upstream interactions and are influenced by organizational context, inter-organizational dynamics, and temporality, which can lead to calibrated or uncalibrated trust and inappropriate reliance. These insights point to new governance challenges and research opportunities for defining trustworthy LLM ecosystems and for developing metrics and policies that account for complex inter-actor trust across the supply chain.

Abstract

With the widespread proliferation of AI systems, trust in AI is an important and timely topic to navigate. Researchers so far have largely employed a myopic view of this relationship. In particular, a limited number of relevant trustors (e.g., end-users) and trustees (i.e., AI systems) have been considered, and empirical explorations have remained in laboratory settings, potentially overlooking factors that impact human-AI relationships in the real world. In this paper, we argue for broadening the scope of studies addressing `trust in AI' by accounting for the complex and dynamic supply chains that AI systems result from. AI supply chains entail various technical artifacts that diverse individuals, organizations, and stakeholders interact with, in a variety of ways. We present insights from an in-situ, empirical study of LLM supply chains. Our work reveals additional types of trustors and trustees and new factors impacting their trust relationships. These relationships were found to be central to the development and adoption of LLMs, but they can also be the terrain for uncalibrated trust and reliance on untrustworthy LLMs. Based on these findings, we discuss the implications for research on `trust in AI'. We highlight new research opportunities and challenges concerning the appropriate study of inter-actor relationships across the supply chain and the development of calibrated trust and meaningful reliance behaviors. We also question the meaning of building trust in the LLM supply chain.

An Empirical Exploration of Trust Dynamics in LLM Supply Chains

TL;DR

This paper broadens the study of trust in AI by examining trust dynamics across the entire LLM supply chain rather than limiting analysis to the user–LLM dyad. Through in-situ qualitative interviews with 71 practitioners across 10 organizations, it reveals a diverse set of trustors and trustees—including humans and various technical artifacts—and documents how trust cues based on the ABI framework (ability, benevolence, integrity) vary across actors. The findings show that trust relationships span downstream and upstream interactions and are influenced by organizational context, inter-organizational dynamics, and temporality, which can lead to calibrated or uncalibrated trust and inappropriate reliance. These insights point to new governance challenges and research opportunities for defining trustworthy LLM ecosystems and for developing metrics and policies that account for complex inter-actor trust across the supply chain.

Abstract

With the widespread proliferation of AI systems, trust in AI is an important and timely topic to navigate. Researchers so far have largely employed a myopic view of this relationship. In particular, a limited number of relevant trustors (e.g., end-users) and trustees (i.e., AI systems) have been considered, and empirical explorations have remained in laboratory settings, potentially overlooking factors that impact human-AI relationships in the real world. In this paper, we argue for broadening the scope of studies addressing `trust in AI' by accounting for the complex and dynamic supply chains that AI systems result from. AI supply chains entail various technical artifacts that diverse individuals, organizations, and stakeholders interact with, in a variety of ways. We present insights from an in-situ, empirical study of LLM supply chains. Our work reveals additional types of trustors and trustees and new factors impacting their trust relationships. These relationships were found to be central to the development and adoption of LLMs, but they can also be the terrain for uncalibrated trust and reliance on untrustworthy LLMs. Based on these findings, we discuss the implications for research on `trust in AI'. We highlight new research opportunities and challenges concerning the appropriate study of inter-actor relationships across the supply chain and the development of calibrated trust and meaningful reliance behaviors. We also question the meaning of building trust in the LLM supply chain.
Paper Structure (25 sections, 1 figure)

This paper contains 25 sections, 1 figure.

Table of Contents

  1. Introduction
  2. Brief Overview of the Background and Methodology
  3. Preliminary Findings & Discussion
  4. Trustors & Trustees: Trust Relationships Are Extremely Diverse Across the Supply Chain
  5. Many Trustors and Trustees Shape the Supply Chain
  6. Human trustors and trustees can be organizational or individual entities
  7. Individual and organizational entities play diverse roles vis-a-vis the LLM, entailing different trust activities
  8. Trusting LLMs is more than trusting a stand-alone, shrinkwrapped, technical artifact
  9. Many Trust Relations Connect These Entities Across the LLM Supply Chain
  10. Downstream to Upstream, Entities Trust Different Technical Artifacts
  11. Upstream Entities Trust Downstream Entities to Responsibly Use Technical Artifacts
  12. Discussion
  13. Trustee-Related Factors: Trust is Fostered By Diverse Trust Cues About the Trustee
  14. The ABI Properties of Organizations Target Trustors Differently
  15. Not All Individual Trustees Display All ABI Properties
  16. ...and 10 more sections

Figures (1)

  • Figure 1: The primary entities (boxes) and trust relations (arrows) that shape the supply chain. The cross and loop arrows respectively represent trust relations across different entities in the same box (e.g., developer and deployer organization), and trust relations across entities of the same type (e.g., two different developer organizations). Remember that the supply chain is not only made of trust relations, but also other relations across entities (e.g., one developer organization might develop a pre-trained model and another one might adopt this model to fine-tune the LLM.)