Table of Contents
Fetching ...

Designing Human-AI Systems: Anthropomorphism and Framing Bias on Human-AI Collaboration

Samuel Aleksander Sánchez Olszewski

TL;DR

It is demonstrated that cognitive biases can impact human-AI collaboration and the need for tailored approaches to AI product design, rather than a single, universal solution, is highlighted.

Abstract

AI is redefining how humans interact with technology, leading to a synergetic collaboration between the two. Nevertheless, the effects of human cognition on this collaboration remain unclear. This study investigates the implications of two cognitive biases, anthropomorphism and framing effect, on human-AI collaboration within a hiring setting. Subjects were asked to select job candidates with the help of an AI-powered recommendation tool. The tool was manipulated in a 3 x 3 between-subjects design to present three different AI identities (human-like, robot-like, generic) and three types of framing (positive, negative, and neutral). The results revealed that the framing of AI's recommendations had no significant influence on subjects' decisions. In contrast, anthropomorphism significantly affected subjects' agreement with AI recommendations. Subjects were less likely to agree with the AI if it had human-like characteristics. These findings demonstrate that cognitive biases can impact human-AI collaboration and highlight the need for tailored approaches to AI product design, rather than a single, universal solution.

Designing Human-AI Systems: Anthropomorphism and Framing Bias on Human-AI Collaboration

TL;DR

It is demonstrated that cognitive biases can impact human-AI collaboration and the need for tailored approaches to AI product design, rather than a single, universal solution, is highlighted.

Abstract

AI is redefining how humans interact with technology, leading to a synergetic collaboration between the two. Nevertheless, the effects of human cognition on this collaboration remain unclear. This study investigates the implications of two cognitive biases, anthropomorphism and framing effect, on human-AI collaboration within a hiring setting. Subjects were asked to select job candidates with the help of an AI-powered recommendation tool. The tool was manipulated in a 3 x 3 between-subjects design to present three different AI identities (human-like, robot-like, generic) and three types of framing (positive, negative, and neutral). The results revealed that the framing of AI's recommendations had no significant influence on subjects' decisions. In contrast, anthropomorphism significantly affected subjects' agreement with AI recommendations. Subjects were less likely to agree with the AI if it had human-like characteristics. These findings demonstrate that cognitive biases can impact human-AI collaboration and highlight the need for tailored approaches to AI product design, rather than a single, universal solution.
Paper Structure (22 sections, 9 figures, 2 tables)

This paper contains 22 sections, 9 figures, 2 tables.

Figures (9)

  • Figure 1: Own elaboration based on Google Trends data for the interest in the term ‘AI’ from 12/29/2019 to 01/27/24. Numbers represent search interest relative to the highest point. A value close to 100 indicates that we are currently at the peak popularity of the term.
  • Figure 2: Slide of a candidate from the training section with its three main sections: Academics, Experience, and AI Recommendation. Example of a subject exposed to the neutral condition of both treatments.
  • Figure 3: Example of one of the two attention checks. Although the structure of the slide remains the same, the question changes. Instead of the question: “Do you invite the candidate for an interview”, subjects are now asked “Did Corina specialize in Arts during Higher Secondary Education?"
  • Figure 4: An example of a neutral frame with positively framed information on the first column.
  • Figure 5: An example of a neutral frame with negatively framed information on the first column.
  • ...and 4 more figures