Abstract

The development of artificial intelligence (AI) has opened a new debate about the linguistic capabilities of this technology and its potential impact on all dimensions of human activity (Brynjolfsson and McAfee, 2014). Aspects such as the massive collection of data to train AI models have raised concerns about the privacy and security of information, due to the risk of misuse. In order to assess similarities or differences between AI and human linguistic abilities, this work specifically analyses argumentative behaviour in terms of the viewpoints adopted by Chat GPT, version 3.5, in the face of a controversial situation expressed in a moral dilemma of the D.I.T.’ questionnaire (Rest, 2013) and is compared with a similar sample of human adults (Noemi, 2019) subjected to the same discursive task. The research followed a comparative multiple case study (Flick, 2020) through a mixed methodology (Johnson and Onwuegbuzie, 2004), combining a quantitative approach with a qualitative one. The results show that 100% of the IH sample take a position (for or against), whereas only 23.4% of the GPT chat sample do. The results encourage us to think that the differences found could be explained by the AI’s weakness in context sensitivity (lack of relevant pragmatic competence) and the lack of a self-involvement process to face moral reasoning tasks.