Generative AI is transforming the way people study, do research, write and carry out professional work. It’s created new opportunities for students, educators, researchers and professionals alike. It has also prompted a host of questions around accuracy, academic integrity, privacy and accountability. Can users tell the difference between good information and bad AI generated content? Do they appreciate the risks of using AI produced material? Can they explain and justify the work they’ve done with the aid of AI? These are questions that are being asked by universities across the globe.
The University of Alabama launched an initiative to deal with these issues via practical AI literacy training. This is the UA AI Experience, a free online course which will take around three hours and is self-paced. It provides the basics of AI literacy but also includes several practical exercises where you are required to assess AI-generated content and make your own decisions. The key point is that effective use of AI is predicated on human judgement, careful checking and professional responsibility.
Building AI Literacy Across the University
The UA AI Experience was developed by the Office of Teaching Innovation and Digital Education in partnership with the UA Teaching Academy. It was launched for the entire university community in Fall 2026 and is open to participants from all academic and professional areas. The approach acknowledges that generative AI is now being used in research, teaching, learning, writing and in all sorts of admin work. Students are automatically enrolled via Blackboard, faculty and staff register to take part. It’s an online experience so participants can complete it at their own pace.
The course is made up of five modules which blend the theory with the practice. The first four modules are all about building up the core competencies and the fifth module is where you get to put that knowledge into practice via a chosen learning pathway. You’ll be doing interactive lessons, looking at sample responses and assessing the quality of the AI output. You’ll be reflecting on your own decisions and thinking about whether the output would be suitable for an academic or professional context. It’s all about linking AI literacy to the actual situations that people will face in their day to day work.
At the heart of the programme is the 4D Model (Discern, Direct, Double Check, Defend). This is the model that provides the structure for the use of generative AI. Discern is about working out whether or not you should use AI in the first place. Direct is about giving very clear instructions and appropriate context to get a useful response. Double Check is about checking the accuracy of the output, checking the evidence, looking for bias. Defend is about explaining your own decisions and being accountable for any work that you’ve done with AI. That’s the 4D model in a nutshell and it provides a very practical framework for the responsible use of AI.
The model is very clear that AI is there to augment human reasoning not to supplant it. Before you even go to the stage of using the AI tool you’re expected to think about the purpose of the task and the risks involved. You’re also expected to consider whether any sensitive data should be stripped out of the prompt. Once you have the AI response you’re expected to assess the quality of that response rather than just go with it. The onus of responsibility always lies with the individual who decides to use it or to share it.
Learning to Question AI-Generated Information
One of the most useful lessons of the course is around the reliability of AI-generated content. AI can produce very convincing explanations, statistics and references. The point being that just because something is well written doesn’t mean it’s true. The course shows how a claim can be entirely unsubstantiated and yet appear perfectly credible when it’s presented in a confident manner with a logical structure. The take away is that AI generated citations are just leads and must be checked independently before they can be treated as evidence.
A practical example of this is an activity which poses the following claim: a study has shown that students are 2.4 times more likely to retain information if they are actively involved in the learning process. The participants are then asked to assess whether the study actually exists and whether the evidence would support the figure given. They are encouraged to go and find the original source and check the evidence before they would ever use such a claim in any presentation. Very much relevant to researchers.
The course also looks at the potential for AI-generated answers to be biased or to contain faulty reasoning. One exercise is to have participants describe a successful entrepreneur. The prompt is then altered to specify that the entrepreneur is from a rural background or that they are the child of immigrants. The way the response is generated can be affected by these changes. The participants learn that the way a prompt is worded can affect the nature of the response and the perspectives that are highlighted. They are encouraged to be sceptical, to look for missing perspectives and to assess whether the response provides a balanced account.
Applying AI Literacy Through Practical Activities
Module 5 provides the opportunity to apply the full 4D Model to real world academic and professional tasks. You could be working with professional emails, project checklists, meeting summaries, academic explanations, practice questions, written paragraphs and arguments. The task is to assess whether the AI output is accurate, relevant and appropriate for the task in hand. You also have to decide whether the output can be used as is or whether it needs further work.
A meeting summary exercise is a good illustration of the risks involved in AI-assisted professional communication. The participants are given a transcript of a meeting which covers the project launch, training, website, and technical access. There are a number of issues still outstanding and no responsibilities have been assigned. The task is to assess whether the AI can summarise the discussion without presenting any tentative ideas as definite decisions. You’re expected to spot the assumptions that haven’t been supported, the information that’s missing and the responsibilities which were never actually agreed.
Other activities are concerned with academic writing and learning. Participants are asked to assess whether AI can improve the clarity of a paragraph without in any way altering the author’s original meaning. They are also asked to consider whether the suggested changes improve the organisation of the ideas whilst maintaining the key links. The learning activities involve looking at AI generated explanations and practice questions to assess their accuracy, relevance and appropriate level of difficulty. This shows that improved readability is not in itself a sufficient outcome if the important information or the intended meaning has been lost.
The activities always come back to the question of whether the user would be prepared to put their name to the AI-assisted work. It’s not simply a matter of checking the grammar or just accepting the content as being ‘okay’ in terms of being professional. The user has to decide whether the response is a true reflection of the original and whether it actually meets the requirements of the task. Where weaknesses are still present they are told to go back and amend the work or re-think their approach. Human judgement is therefore a key part of AI-assisted work.
Recognizing Achievement and Professional Responsibility
The UA AI Experience utilises digital badges to acknowledge participants’ progress and achievement. There are four core badges which correspond to the key learning competencies: Discernment Certified, Director’s Cut, Verification Pro and Defender of the Standard. To be eligible for the official UA AI Experience completion badge, participants must complete the four foundational modules plus a required fifth module pathway. The badges serve as a record of participation and can be shared via professional networks.
I did the course activities (October 22-23 2026) and received the digital achievement badges via the uni’s learning platform. As someone in an academic role (teaching and research) I found the activities very relevant to the work we do in terms of writing and communication. The modules on citation checking, reasoning, bias and accountability were all very useful. They really drove home the point that you do need to check any information generated by AI before you use it in your research, in your teaching or in any academic publication.
The experience also underlined why AI literacy needs to go beyond just knowing how to construct good prompts. You can quite easily generate a response that appears to be of value and yet contain factual errors or make assumptions that are entirely unsupported. Responsible use of AI requires you to be critical of both the information being provided and the ramifications of that information. The course was designed to get people to think about those issues for themselves rather than simply accept that the AI had done its job and the output was therefore correct.
Lessons for Universities Beyond the United States
The University of Alabama’s example is one that other higher education institutions could look at when designing their own AI literacy programmes. It’s a very short, self-contained programme which introduces the basic principles without any requirement for a high level of technical expertise. The practical elements of the programme also mean it is relevant to a wide range of disciplines across the university, from business through to education, the social sciences, the humanities and so on. Students and staff in those various disciplines will all be likely to face similar issues relating to the use of AI, including the potential for incorrect information, for biased explanations and for inappropriate use.
For universities in Bangladesh and in other developing higher education systems the programme offers some very useful ideas for institutional AI training. It could form the basis of a programme to train staff, to train students and to support research. Of course any such initiative would have to be tailored to the local context, to the actual requirements of the institution, to the level of digital access and the level of prior experience with AI that would be available. It would not be a case of just taking the course and running with it.
The model could be of considerable use in establishing a common understanding of responsible AI use across an organisation. It could be applied by staff in the faculty when assessing AI-assisted academic writing, by students when assessing the information they are using for their own work, by researchers in their own use of AI to support research and by admin staff in their use of AI in relation to privacy, communication and accountability. All of these would of course require appropriate training and evaluation to determine whether they are effective.
However doing a short AI literacy course doesn’t in itself lead to responsible use or to any kind of long term competence. Ongoing practice, institutional support and regular review are all still required. Any university looking at such a course would be well advised to consider whether the course has actually led to an improvement in participants’ ability to spot errors, to protect sensitive data and to make good decisions. There’s also scope for research into the effect of such training on actual practice over time.
About me:
Dr Khandakar Kamrul Hasan is an Associate Professor in the Department of Business Administration at the University of Scholars, Bangladesh. He completed his first PhD in Business Management at Tripura University (a Central University), India. He is currently doing his second PhD in Higher Education at the College of Education, The University of Alabama, USA. He completed the University of Alabama's AI Experience in October 2026.
