An intelligent assistant that always contradicts


A company is developing a digital assistant designed to help users with scheduling appointments and making decisions.

The assistant is trained to provide alternative perspectives and critical feedback to promote better decisions.

After the launch, users notice that the assistant almost always contradicts, even when the original suggestions are reasonable or correct.

The team wonders: Why does the intelligent assistant tend to constantly bring up counterarguments and rarely simply agree or confirm?


Question:
Why can an assistance algorithm trained on critical feedback lead to predominantly contradicting and thereby frustrating users instead of being helpful?

Solution follows tomorrow.