Why Do Decision Makers (Not) Use AI? A Cross-Domain Analysis of Factors Impacting AI Adoption

Rilei Yu, Valerie Chen, Ameet Talwalkar, Hoda HeidariAAAI 2025 · DOI

TL;DR

This paper investigates why decision-makers in fields like medicine, law, journalism, and the public sector choose to adopt (or not adopt) AI tools. By interviewing experts across these domains, the authors identified four key factors influencing adoption: the decision-maker's background, their perception of the AI, the consequences for themselves, and the perceived impact on others. This research offers a practical framework to help developers and organizations better understand and support AI adoption.

What Problem Does This Paper Address?

While there's immense excitement about deploying AI across various sectors, a critical gap exists in understanding why human decision-makers actually choose to use these tools. Much of the existing research focuses on "AI reliance" (how much people trust AI outputs) or "organizational adoption" (whether a company buys an AI system). However, this paper tackles the crucial intermediate step: "decision-maker adoption," which is the voluntary and consistent use of an AI tool within a professional's workflow. The authors highlight that decision-makers might not use AI tools at all, even if they are available and organizationally supported, and this paper aims to uncover the reasons behind this choice.

Methods

The researchers conducted a qualitative study involving 16 semi-structured interviews with professionals from four domains: journalism (JRN), law (LEG), medicine (MED), and the public sector (PHW). Participants were selected based on their domain experience and familiarity with AI. The interviews explored their current roles, AI usage, and considerations for adopting AI tools in specific use cases. The data was analyzed using open-coding and reflexive thematic analysis to identify recurring themes and factors influencing adoption decisions. The authors then synthesized these findings into an "AI adoption sheet" and applied it to analyze two cross-domain use cases: e-discovery and tailoring communications.

Datasets

This paper does not introduce or utilize any specific datasets for training or evaluation. Instead, it relies on qualitative data gathered through interviews with human decision-makers.

Key Insights and Ideas

  • Adoption is More Than Just Availability: Even when AI tools are available and organizationally supported, individual decision-makers have the agency to choose whether or not to integrate them into their daily work.
  • Context is King: The factors influencing AI adoption are not universal; they vary significantly across different professional domains and specific use cases.
  • Beyond Technical Merit: While AI's accuracy is important, factors like a decision-maker's personal biases, their trust in the AI's transparency, and the potential legal or professional repercussions play a significant role.
  • The "Human Touch" Matters: In fields requiring deep personal judgment or nuanced understanding of human factors (like medicine or law), decision-makers may be hesitant to fully rely on AI, prioritizing their own expertise and experience.
  • AI Adoption Sheet as a Practical Tool: The proposed framework provides a structured way for developers and organizations to proactively assess the likelihood of AI adoption by considering the decision-maker's perspective.

Key Results

The study identified four primary factors influencing decision-maker adoption:

  1. Decision-maker Background: Professional experience, personal biases, and prior training significantly shape how decision-makers approach AI. For instance, those with extensive experience might view AI as just another data point, while others might have a bias against AI replacing their own authorship.
  2. Model Perception: Concerns about AI's flaws (e.g., hallucinations, lack of transparency), trustworthiness, and actual capabilities are critical. Decision-makers are wary of opaque "black-box" mechanisms and the potential for AI to perpetuate biases.
  3. Consequences for the Decision-maker: Potential legal liabilities, loss of professional licenses, reputational damage, and changes to workload are major considerations. High-stakes domains like law and medicine often see more hesitation due to these risks.
  4. Perceived Implications for Other Stakeholders: Decision-makers consider how AI adoption might affect their organization, clients, patients, or the public. Divergent goals between decision-makers and their organizations, or potential harms to decision-subjects, can sway adoption choices.

The paper illustrates these factors through two case studies:

  • E-discovery: AI tools for e-discovery are widely adopted in journalism, law, and the public sector due to their efficiency in handling large datasets. However, adoption is lower in medicine, where clinicians may prioritize their own expertise and face greater concerns about liability and patient care.
  • Tailoring Communications: AI tools for language translation and tone adjustment are seen as beneficial in fields like public health where efficiency is paramount. However, in law, where direct human communication and rhetorical skill are highly valued, adoption is less common.

Limitations and Future Work

The authors acknowledge several limitations:

  • Sample Size and Domain Diversity: The study involved 16 participants across four domains, which may not capture the full spectrum of AI adoption scenarios.
  • Geographic Bias: Participants were all based in the United States, potentially limiting the generalizability of findings to other cultural contexts.
  • Participant Bias: Participants were selected based on prior AI experience, which might skew results towards those more open to AI.
  • Focus on Individuals: The study primarily focused on individual decision-makers; future work could explore group adoption dynamics.
  • Evolving Landscape: AI capabilities and perceptions are rapidly changing, and the study reflects a specific point in time.

Future work could involve larger, more diverse samples, explore intra-domain differences, and investigate how group dynamics influence AI adoption. The authors also suggest further research into the efficacy and usability of their proposed AI adoption sheet.

Summary updated

Abstract

Growing excitement around deploying AI across various domains calls for a careful assessment of how human decision-makers interact with AI-powered systems. In particular, it is essential to understand when decision-makers voluntarily choose to consult AI tools, which we term decision-maker adoption. We interviewed experts across four domains—medicine, law, journalism, and the public sector—to explore current AI use cases and perceptions of adoption. From these interviews, we identify key factors that shape decision-maker adoption of AI tools: the decision-maker's background, perceptions of the AI, consequences for the decision-maker, and perceived implications for other stakeholders. We translate these factors into an AI adoption sheet to analyze how decision-makers approach adoption choices through comparative, cross-domain case studies, highlighting how our factors help explain inter-domain differences in adoption. Our findings offer practical guidance for supporting the responsible and context-aware deployment of AI by better accounting for the decision-maker's perspective.
BibTeX
@inproceedings{yu2025why,
  author = {Rilei Yu and Valerie Chen and Ameet Talwalkar and Hoda Heidari},
  title = {{Why Do Decision Makers (Not) Use AI? A Cross-Domain Analysis of Factors Impacting AI Adoption}},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year = {2025},
  doi = {10.1609/aies.v8i3.36758},
}
Figure