Table 3 Pro and contra reasons for using AI-driven patient preference prediction

Increased Speed

De-Humanization

 ■ “Time saving with quick evaluation of a situation”

 ■ “A decision based on demographic data is precisely not an individual one”

 ■ “Faster decision-making, e.g. in triage situations”

 ■ “Too much generalization – too little attention paid to individual patient needs”

 ■ “Rapid availability in borderline situations without the possibility to wait for ethics consultation or discussion among colleagues”

 ■ “Statistical programs run the risk of only depicting the peak of the Gaussian distribution”

Ensured Objectivity

Overreliance

 ■ “Possibility of an objective collection of the data”

 ■ “Blind and unreflected trust in the program”

 ■ “Assessment outside my medical goals and philosophy of life/own quality of life”

 ■ “The decisions of a computer cause a false sense of security for the decision-makers and can secondarily lead to conflicts of conscience”

 ■ “Independence from ethical currents and schools or religious norms”

 ■ “What I see above all is the risk that the program will be presumed to be responsible and one ceases to think for themselves and to deal with the patient and his or her individual situation”

Procedural Benefits

Incorrect Predictions

 ■ “Suggestions from the AI can support the decision-making process”

 ■ “Decisions that are too complex for an AI can lead to incorrect advice”

 ■ “AI could, through automation and standardization, identify and present aspects that would otherwise not be considered”

 ■ “Potentially worse/less accurate decision than by the doctor”

 ■ “Additional component and further perspective to be able to put the decision on a broad basis”

 ■ “A machine could come to wrong conclusions in borderline cases”