One striking aspect of the new Ethical Framework for AI in Education is how little of it is actually about AI technology. The Framework has nine objectives and 33 criteria: 18 of these apply to the ‘pre-procurement’ stage, and another five to ‘monitoring and evaluation’.
That’s a refreshing change from the usual technology-led discussions in this space: here it’s almost all about the organisation within which the AI will work. Do we understand our goal in choosing to use AI, is there a sound educational basis for that, what changes will this involve for processes and skills, do we understand the risks, how can we detect and change course if it doesn’t work out? And, equally important, does our supplier understand what we are trying to achieve and commit to supporting our choice of goals and assessment of risks?
Even the seven ‘implementation’ criteria are about process: how can AI be used in assessment to demonstrate skills and support well-being; how can we create safe spaces outside continuous assessment; how can AI help us avoid unfavourable outcomes for individuals; how will we help all stakeholders (students and staff) work effectively and ethically with AI; how will we manage the changes that introducing it should bring?
With this comprehensive understanding of the context we want AI to support and enhance, the actual technology choice should be much simpler. Some technologies (maybe even some applications) will be clearly unsuitable: others will be a good, or perfect, fit. Best of all, we’ll be able to provide the most important explanation for trustworthy AI: why we chose to use it.
