Moral considerations relating to the use of ChatGPT

Our project saw a two hour workshop developed and delivered to first and second year undergraduate music, theology and philosophy students in DCU (staff were invited to attend). . Part of the workshop was lecture-based. The lecture focused on the architecture of neural networks, the way networks are trained, and the design of LLMs and ChatGPT in particular. The question whether ChatGPT is exhibiting “understanding” or not was also addressed (albeit at a basic level). In the second part of the workshop, the students engaged in group discussions on the topics of ‘transparency’, ‘bias’, ‘fairness’ and ‘environmental cost’. These topics were not theoretically defined beforehand: students' own understanding of these terms was the starting point of the discussion. Groups were led by facilitators (Philosophy staff members) who partly followed the Socratic method; e.g., not providing answers but asking questions that help students analyse their beliefs and uncover hidden assumptions (the Socratic method is an effective method in moral education; see, e.g., (Pihlgren, 2008); (Torabizadeh et al., 2018)). A handout with further data on environmental costs and some real-life examples to discuss was made available to help discussion along when needed.  We also delivered a shortened version of the workshop to all third year BAJH philosophy students and some students from the DCU Institute of Education. 

On the back of the workshops we are developing an interactive tutorial on the same topics. We use 5HP software to provide a level of interactivity and incorporated slides with narration. A research associate from the DCU computer science department is helping  to develop illustrations. The final version of the tutorial will be made available as an open educational resource.  

Conference talks informed by our project experiences were given at the European Conference on Ethics and Integrity (ECEIA 2026 -  June 2025, Uppsala University), the UCD College of Social Sciences and Law Research Day (March 2026, UCD), and Meaning and Mind in AI, SOPHIA 2026, Symposium on Philosophy & Integrity in Healthcare AI (30th of April 2026, University College Dublin). In addition various talks related to Teaching and Learning were given, e.g. as part of the Digital GenAI and Teaching Badge course, DCU’s Teaching and Learning week and the National Forum Associates meeting.   

Student engagement and learning: Students were positive about the workshop and there were indications that workshop attendance had an effect on intended future academic and leisurely use of LLMs. We applied and received ethics approval to conduct anonymous surveys (pre and postworkshop) and a follow up survey one month later to future participants in the workshops in order to  investigate whether increasing higher education students’ understanding of how LLMs are designed and trained - along with encouraging reflection on the ethical implications of this - decrease the (possibly illicit) use of LLMs in students' coursework. 

Curriculum design: The development of the workshop and the tutorial informed the design of a new Philosophy of Mind and AI module that will be incorporated into the BAJH curriculum from September. 

Teaching and learning approaches: On the back of the workshop experience, we gave a presentation to the Faculty of Arts, Humanities and Social Sciences at the University of Limerick which was meant to partly inspire discussion on assessment that same day. We also gave a presentation to academic staff at the DCU School of Theology, Philosophy, and Music on the latest developments in genAI, and some staff reported that this helped their thinking on assessment. 

The reception of the workshops and draft tutorial brought home to us the importance of AI literacy (in its broadest sense). As mentioned above, we have developed a new Philosophy of Mind and AI module, partly informed by our experiences. We would also like to support other initiatives to introduce modules on AI literacy in the different  Schools (other than the School of Computing, of course).  In the meantime we will continue to offer workshops on these matters. 

As far as our own practice with regard to assessment is concerned, some of us have started to include in-class exams as a midterm examination so that the emphasis on an end-of-term essay (typical for philosophy students) is reduced. 

What really was driven home to us was the high level of interest of students in how LLMs work and the willingness to think deeply about possible ethical issues relating to the design and the use of LLMs.  Some students embraced the use of AI and others stayed away from it but students were keen to be fully informed to make these decisions. We are firmly of the opinion that giving the required information and letting students make their own decisions works better than either promoting or banning use of LLMs (although use should of course be within the bounds of  academic integrity).