A Study on Identifying Human Factors in Collaboration with the AI-infused Systems

AI technologies are spreading into many fields, and as interaction between humans and AI grows, difficulties and errors can arise. The paper notes that users may interact differently with AI depending on human factors, which creates challenges in human–AI relationships. It argues that human factors (human behavior, abilities and limitations applied to system design) should be studied to optimize performance, ensure safety and reduce errors. The paper aims to analyze human factors in AI and discuss human–AI collaboration based on prior research in healthcare, education and automated driving.

A literature-based study in three steps (Table 1):

1. Literature review of articles on human factors and AI in autonomous driving, education and healthcare.
2. Human factor analysis: factors were grouped into categories (demography, physical capabilities, health, cognitive, experience, behavior, social/cultural influence, technology trust) and tallied by field and by the number of studies that used them.
3. Summary: selection of the most critical factors to consider when using AI-infused systems.

• Review finding: gender (19 studies) and age (18 studies) were the most frequently cited factors and appeared in all three fields; trust in ADAS/technology (12 studies) also appeared in all three.
• Review finding: technological expertise (9 studies) and social/cultural influence, namely family (5) and collectivism (6), were important in education and healthcare.
• Review finding (autonomous driving): besides demographics, cognitive and physical factors such as focus, multitasking, IQ, learning speed and vision were reported to affect intention to use automated vehicles or ease of adaptation. The factor table also lists experience-related factors for this field only, such as license type and ADAS exposure years.
• Review finding (education and healthcare): in education, school location and teachers' residential area may affect adoption, since rural teachers use less modern technology. In healthcare, initial trust increased professionals' willingness to adopt an AI-based diagnosis support system.
• Review finding (research gap): autonomous driving is the most discussed field, while research on AI adoption in education and healthcare that considers differences in health, cognitive, work and physical capabilities is still lacking.

The paper concludes that identifying human factors is highly important when applying AI in the three fields. It states that considering these factors in AI system design increases ease of use, speeds adaptation to the technology and reduces potential errors. It attributes the smaller body of research in education and healthcare to the lower use of smart technologies there, and notes that age, gender and trust in technology are equally important across the three fields. The authors suggest the findings can support further studies on the causes of human error in takeover or handover scenarios with AI-infused systems. The paper does not state limitations.