A Survey of AI Reliance

Sven Eckhardt,Niklas Kühl,Mateusz Dolata,Gerhard Schwabe

2024-07-22

Abstract:Artificial intelligence (AI) systems have become an indispensable component of modern technology. However, research on human behavioral responses is lagging behind, i.e., the research into human reliance on AI advice (AI reliance). Current shortcomings in the literature include the unclear influences on AI reliance, lack of external validity, conflicting approaches to measuring reliance, and disregard for a change in reliance over time. Promising avenues for future research include reliance on generative AI output and reliance in multi-user situations. In conclusion, we present a morphological box that serves as a guide for research on AI reliance.

Human-Computer Interaction,Artificial Intelligence,Computers and Society

What problem does this paper attempt to address?

The paper aims to investigate and summarize existing research on artificial intelligence (AI) dependency and propose future research directions. The authors point out that while AI systems have become an indispensable part of modern technology, understanding how humans rely on these systems is relatively lagging. Specifically, the current research has the following issues:

1. **Unclear factors influencing AI dependency**: It is currently unclear which factors affect the degree to which people rely on AI recommendations.

2. **Lack of external validity**: Existing studies are often conducted in specific contexts, making it difficult to generalize to other practical application scenarios.

3. **Inconsistent methods of measuring dependency**: Different studies use different methods to measure dependency, leading to issues of comparability between results.

4. **Neglect of changes in dependency over time**: Few studies focus on whether individuals' dependency on AI systems changes over time.

To address these issues, the paper provides the following contributions through a literature review:

- **Introduction of a socio-technical perspective**: Emphasizes the need to consider AI dependency from multiple angles, including the interaction between technology and human users and the broader social environment.

- **Identification of research limitations**: Points out the shortcomings in existing research, providing guidance for future related studies.

- **Proposing new research directions**: Based on current research gaps, proposes possible future research paths, such as dependency on generative AI outputs and dependency in multi-user scenarios.

- **Providing a classification framework**: Designs a classification framework that can be used to analyze and guide AI dependency research.

In summary, the goal of this paper is to provide a systematic understanding framework for AI dependency research and offer practical tools and suggestions for researchers to promote further development in this field.