Artificial Intelligence and “Decision Intelligence”

Letters to the Editor regarding the article The Quality and Utility of Artificial Intelligence in Patient Care by Prof. Dr. med. Kai Wehkamp, Prof. Dr. med. Michael Krawczak, and Prof. Dr. med. Dr. h.c. Stefan Schreiber in issue 27–28/2023

The authors touched on important aspects of using artificial intelligence (AI) in patient care (1). The explained limitations and risks—such as bias at various levels, socioeconomic aspects, overfitting and underfitting, and others—represent a partial aspect of so called decision intelligence. This discipline combines insights from statistics, data sciences, social sciences, and decision theory. It aims to generate better decisions from data. In my own experience, knowledge of statistics, probability theory, and data science make the results of AI models more tangible. To be able to recognize the different forms of cognitive biases make handling AI easier, but not only that. Cassie Kozgyrov, Google’s head of decision intelligence, has published many articles and videos on decision intelligence and machine learning (2). These articles provide a good entry point into these subjects.

The Stratos-Initiative (STRengthening Analytical Thinking for Observational Studies), among others, provides more in-depth information (3). The academic work coming out of this initiative—especially the working group “Evaluating diagnostic tests and prediction models”—provides suggestions for handling prediction models (4).

DOI: 10.3238/arztebl.m2023.0213

Prof. Dr. med. Peter Hahn

Vulipius Klinik, Bad Rappenau

References

1. Wehkamp K, Krawczak M, Schreiber S: The quality and utility of artificial intelligence in patient care. Dtsch Arztebl Int 2023; 120: 463–9

2. Making friends with machine learning: The Entire Course. YouTube Video. www.youtube.com/watch?v=1vkb7BCMQd0&list=PLRKtJ4IpxJpDsOT_8YDREJrO8cQUtPUVg&index=4 (last accessed on 28 September 2023).

3. The STRATOS initiative. https://stratos-initiative.org/en (last accessed on 28 September 2023).

4. de Hond AAH, Leeuwenberg AM, Hooft L, et al.: Guidelines and quality criteria for artificial intelligence-based prediction models in healthcare: a scoping review. NPJ Digit Med 2022; 5: 2