is a better approach to AI
because it places people at the center of the process. In this manner people may
reach conclusions while understanding how they got there. Our approach to AI has
turned out to have important advantages.
With conventional AI tools the goal has been to reach conclusions
while leaving people out of the process as much as possible. It is
unavoidable that such AI will be positioned to replace people, but this is an
unstable situation given the current state of the art. A symptom of this
instability is the explainability problem.
As an example of the explainability problem, consider what happens when a
person, versus what happens when an AI, scores a bank loan application, or makes a
hiring decision. A person will take into account laws and conventions that are
designed to safeguard against certain types of prejudicial or otherwise known to
be flawed thinking. In contrast, such flawed thinking might be exactly what a
conventional AI program did before reaching an unexplained decision. We just
can not know, because the conventional AI can not explain to us what it did.
Copyright © 2020 by Reasoning Technology. All Rights Reserved.