Drawing a line in the sandbox: balancing exploration and instruction in AI playgrounds

Abstract

Following the release of ChatGPT and similar generative artificial intelligence (GenAI) tools, many universities have committed to helping their students become AI literate (see for example Russell Group, 2023). The means of providing this support vary, ranging from optional, asynchronous guides to taught instruction embedded within courses. One popular format is the ‘AI playground’: practical sessions encouraging students to experiment with different AI tools, often with minimal instruction given by the facilitator.

Yet how much instruction is too much or too little? Some leading voices in AI pedagogy place an emphasis on exploration and self-discovery, encouraging students to set their own goals and play with AI tools rather than walking them through specific tasks (Mollick, 2023). Minimising instruction, however, may sometimes clash with students’ needs. Recent research indicates that many students feel overwhelmed and anxious about AI, and a lack of confidence may hold some back from simply diving in on their own terms (Tierney and Peasey, 2023). 

In order to find a balance between exploration and instruction, this paper presents different formats of AI playgrounds with varying levels of instruction, run by the Study Skills team at the University of Bristol during the 2023/2024 academic year. Using feedback from student attendees, we identify some specific areas of AI use in which students value facilitator instruction, and at what stages the facilitator should minimise their own intervention. With these findings, we present a potential model for introductory AI playgrounds for use and adaptation by other HE practitioners.