Intricacies of human-AI interaction in dynamic decision-making for precision oncology

Area of Science:

• Oncology
• Medical Informatics
• Artificial Intelligence in Medicine

Background:

• AI decision support systems offer personalized cancer care by adapting treatments to patient progression.
• Over-reliance or under-reliance on imperfect AI can result in suboptimal cancer therapeutics.

Purpose of the Study:

• To investigate the Human-AI interaction dynamics in AI-assisted adaptive radiotherapy for non-small cell lung cancer and hepatocellular carcinoma.
• To explore how clinicians collaborate with AI in treatment planning.

Main Methods:

• A Human-AI interaction study was conducted.
• Two levels of collaborative behavior were investigated: model-agnostic and model-specific.
• The study focused on response-adaptive radiotherapy for specific cancer types.

Main Results:

• Human-AI interaction is multifactorial, influenced by clinician's prior knowledge, patient state, disease site, and AI characteristics.
• Clinicians' trust and reliance on AI recommendations vary based on perceived benefit, harm, or model transparency.
• Clinicians may disregard, critically analyze, or adjust decisions based on AI recommendations.

Conclusions:

• Effective Human-AI collaboration in adaptive cancer treatment requires careful consideration of multiple influencing factors.
• Understanding clinician-AI interplay is crucial for optimizing AI's role in personalized cancer care.
• Clinicians' critical judgment remains essential in utilizing AI for therapeutic decision-making.