AI systems using ML/DL are perceived as a “black box”. Neither the doctor nor the patient knows how it works. The authority of the doctor, trust in his knowledge and experience have always played a big role in treatment then patients show anxiety and even hostility towards treatment by "maschine" when their lives and health are at stake. The key is to educate the public and patients.
A possible solution to the “glass box” of artificial intelligence (XAI) (decision tree, linear regression model) will create transparency and build trust in clinical AI.
As far as I see it, AI/ML based systems do have a lot of "limitations", but it does have this amazing ability to perform it's tasks with wonderful accuracy and precision when it is allowed to work within its " limited realm", decades of scientific research have enabled us to develop several quantitative tools to assess the performance of such systems, in terms of the model overfitting issues.
For the most part, AI systems are systematic and sophisticated information extraction and processing systems that to certain extent can mimic the human reasoning and knowledge acquisition process and in certain situations, can outperform the human intelligence too. In order to make the upcoming next generations of human civilization more comfortable/conversant with AI, academic initiatives need to be taken up from young academic levels itself, preferably right from the beginning of the college years or so.
Mr Debopam Ghosh , these are very valid observations about the advantages of AI. However, it is necessary to draw attention to the important psychological aspects of the use of devices with AI algorithms in medicine especially as they relate to such a sensitive area for patients as their health and life. The topic of AI is often sensationalised in the media, and it is necessary to raise public awareness through educational policies, as in relation to the patient .
Hanna Mielniczuk "AI systems using ML/DL are perceived as a “black box”."
Are you certain that this point is still valid overboard? I mean if you run let's say llm with chain of thought or deep search, then you get information on thought process and rationale behind particular recommendation.
"The key is to educate the public and patients."
The situation right now is that such models (except maybe specialist models optimized to solve very narrow task) tend to have rather annoying tendency to hallucinate, where they make up plausibly sounding answer. If anything, right now there tend to be studies that best performance is achieved by people who in the same time use AI but are quite distrustful and skeptically analyze its output.
"hostility towards treatment by "maschine""
Hard to say, for years there had been high share of population asking Dr Google for medical advice, so I'd lean towards option that they would be willing to start to rely on such models even before they are properly tested.
Dear Marcin Piotr Walkowiak , thank you for your comment. Although today most artificial intelligence systems in medicine are black boxes, there is an emerging trend to build Explainable Artificial Intelligence (XAI) systems, especially in medicine. Article Medical artificial intellegence and the black box problem – ...
The conclusion “that the best performance is achieved by people who use artificial intelligence at the same time, but are quite distrustful and skeptical of its results” is in my opinion in agreement with the doctrine of EBM (Sackett, 1997) “The practice of evidence-based medicine means integrating individual clinical knowledge with the best available external clinical evidence from systematic ...”
- Badges
- Science topic
- Similar topics
- Biological Science
- Biochemistry
- Biomolecules
- Nucleic Acids
- RNA
- Small RNA
- microRNA