Table 4 Identified themes and categories
Theme | Category | Subcategory |
|---|---|---|
Before interaction | Certainty in own assessment | Perceived uncertainty as prerequisite for information gathering |
Previous experience leads to direct decision | ||
Expectations of SAAs/LLM | General mistrust | |
Fear of misdiagnosis | ||
System is perceived as superior to oneself (due to lack of own medical knowledge) | ||
Data basis | SAA perceived as trustworthy because of medical knowledge base | |
Concerns over general data basis of ChatGPT | ||
During interaction | Information gathering and information analysis | Gathering information: get more information on ambiguous symptoms |
Analyzing information: narrow down decision-making space | ||
Perceived personalization | Criticism of generic responses | |
Increased trust and reliance through perceived personalization | ||
Feeling of social interaction increased trust | ||
Explainability | Need for uncertainty quantification | |
Lack of explainability in both systems | ||
After interaction | Information integration | Accepting system’s recommendation because it fits knowledge or experience |
Seeking additional information to verify inconclusive recommendations | ||
Difficulties in information integration | Too much information felt overwhelming | |
Contradicting recommendations lead to doctor’s visits and higher urgency |
