?
Цифровой театр абсурда: могут ли нейросети поставить новую научную проблему перед психологией? Кейс-сравнение ChatGPT и DeepSeek
The rapid advancement of artificial intelligence technologies has drawn increasing attention from psychological researchers. While neural networks are being integrated into nearly all domains of human activity, the boundaries of their applicability remain unclear — particularly regarding the originality and practical value of the content they generate. Proponents advocate for their widespread adoption, whereas skeptics raise serious concerns about the scope and validity of their competencies.
This article presents a case study comparing two popular large language models — ChatGPT and DeepSeek. The models were tasked with generating an entirely novel scientific problem in the field of psychology, followed by mutual evaluation of each other's creative output. We propose that the formulation of scientific problems — as a specific form of problem construction—can serve as a promising and reliable method for assessing creativity in both humans and machines. While previous studies have demonstrated that neural network models can outperform humans on certain creativity tests, there remains no universally accepted or methodologically rigorous framework for evaluating AI-generated creativity.
Our study is grounded in the theoretical framework of complicatology, a concept suggesting that the generation of difficulties and problem situations is not only diagnostically and didactically meaningful but also a manifestation of creative potential. In line with this framework, we examined the models’ ability to formulate original psychological problems and subsequently assess each other's responses.
The interaction with the models revealed that they were not only capable of generating novel and theoretically grounded psychological problems along with potential solutions, but also demonstrated elements of self-reflection — such as identifying strengths and weaknesses in their own outputs, offering critical feedback, and even producing metaphorical comparisons. One of the most vivid metaphors to emerge spontaneously from the models was that of the “digital theatre of the absurd,” which we adopted as the title of this article.
This case study highlights the potential of large language models for exhibiting creativity and self-reflective reasoning. It may serve as a foundation for developing future methodologies to assess AI creativity in a more systematic and theoretically informed manner.