Last week I looked into AI in assessment with more of a surface level scope, not knowing too much around this topic initially and pulling mainly from the article, Assessment and AI by WesternU.ca and the Centre for Teaching and Learning. This week I really wanted to reapproach it as I found a very intriguing research article that I recommend reading through called, Al assessment changes human behavior, published on PNAS, which uses a psychological lens and also reveals the impacts of informing vs not informing individuals of the use of AI. There are simply so many layers to this exploration and it felt unrewarding leaving some critical and relevant ones unturned. Let’s uncover why AI assessment might change human behaviour, students’ work, ethics and approaches to learning.

Consequences for Those Taking Assessment and Organizations Conducting Assessment

My main thought or expectation when opening this article was on students mimicking teacher behaviour or use of AI. For example a student experiencing their teacher using AI to help or guide with assessment processes might feel driven or even pressured to copy this and use AI to assist in their own assignment processes through generating ideas, text or editing and revising to achieve a certain level that will be acknowledged by AI’s marking positively and reach its “standard”. This might be to satisfy students’ ability to manage or control as best as possible their own academic outcomes linking to grades, image, future and general success. This could in turn shift testing results, human centered contributions/engagement and cause students to build an overreliance on AI.

Over-reliance on AI / Analytical Responses

AI being used in assessment practices might lead to individuals’ tendency to use AI programs to reach success constantly even outside of education leading to lessened authenticity, creativity and mindfulness in their work if not balanced, done in a meaningful or informed way. One of the article’s studies (study 3) reveals this reaction by assessment takers as the article states they exemplified less general, “creativity , ethical and social considerations, risk-taking, and effort investment.” This is further elaborated upon throughout the article as it states, “current research shows that AI assessment leads people to present themselves as more analytical because they believe that AI particularly values analytical characteristics… (reflecting the “AI assessment effect.”)” For some, in order to achieve this AI version of a “perfect mark” might mean relying on its programs during the work portion/creation itself. This also brings into question what a “perfect mark” really means and if this should look different when human-centered or AI assisted, also reflecting what rubrics should represent and include.

Analytical Priority Belief

Based on the article’s findings, there is a strong “analytical priority belief” around AI programs causing what they call, “the AI assessment effect” as described above. This is not always the most accurate and could lead to inaccurate or lacking results by testers as AI programs have varying goals based most often on the way the system was actually trained / its training data. I also went in and asked ChatGPT4, “Which AI programs are not Analytical driven are they programmed to be this way or how is this controlled?” It solidified the fact that this is due to training but more particularly with modern and large language models (LLM’s) training which prioritize pattern-based reasoning.

Final Takeaway

At the end of the day assessment procedures are meant to effectively evaluate individuals authentic human-centered work, strengths and performance in order to provide valuable feedback and considerations to develop further growth and knowledge. As a student educator I would be very specific and open on the idea of using AI programs in this way or even just testing out different programs’ abilities in creating for instance a layered type of assessment integrating layers of human and AI feedback/marking. I feel it’s extremely important to allow students to express themselves freely especially in their everyday classroom contributions and assessment. If this were lessened by AI integration I would aim to prioritize human based assessment processes or AI that is not purely analytically driven to compare impacts within a classroom and encourage students trust around these experiences.