Experimental methods for assessing the quality of products of AI systems
Project goal
is to objectify and optimise the evaluation of the quality of texts generated by artificial intelligence systems using the achievements of experimental social sciences, which use direct metrics of human behaviour, psycho- and neurophysiology to solve similar problems.
The project creates a completely new approach in the field of artificial intelligence systems evaluation, involving the comparison of human eye movement parameters during the reading of texts generated by humans and various artificial agents to:
- Identifying metrics to discriminate between human-generated texts and artificial intelligence systems.
- Ranking the latter by the degree of closeness to natural texts.
