A not-so-quiet revolution is unfolding in the technology space that has enormous implications for writing and assessment in education. A recent article, and many tweets, have declared the college essay dead. On the surface, it’s the typical over-reach statement that people who chase new ideas like to declare in order to get to the front of the attention conga line. At even a slight bit of inspection, such as writing an essay structure in Moonbeam, and you’re clearly confronted with the realization that things have changed and this marks an inflection point. From here, the technology will only get better. For me, this marks one of three major interactions with technology that forced a realization that things had changed fundamentally (the other two being computers and then the read/write web).
It’s not that GPT3 or ChatGPT write brilliantly. It’s that they write plausibly. Sometimes the output is childish. Sometimes it is brilliant. It’s always fast – a four page essay in a matter of seconds. It feels like interacting with something that is at least intelligent-adjacent.
Given how central writing is to our daily lives, and its role in the college curriculum as a marker of student knowledge and progress, when something gets close to good enough it changes things. Moonbeam can offer a structure and outline of an essay that is better than what many undergraduates produce. With a bit of strategic editing during keyword and outline stages, the quality of the output can increase significantly.
I think it’s now unethical to ask students to write a general topic essay and only assess the end artefact. For rudimentary topics, writing is no longer a needed skill. As a tool for thinking and organizing ideas, it still has value. For assessment in university courses, it’s obsolete.
Universities – ever the slow and plodding organism that enables all manner of progress to pass it by – this should be a “hit the panic button” moment. Time to rethink where actual points of value exist in the educational process. Deans and Chairs should be calling urgent meetings and hosting sensemaking discussions.
Most critically though, they should be focusing on the skills that we should be developing in students in the age of AI. We just finished our third annual conference on Empowering Learners for the Age of AI. The advancements year over year are hard to overstate. The implications for universities are become more dire. For me, the logic is rather simple: if there is an entity that has intelligence-like capability that year over year advances at a pace that vastly exceeds our biological progress, then the nature of that intelligence should give us pause and we should evaluate what and how we teach in education.
A colleague, Abelardo Pardo, stated it nicely:
The way I see it is that so far we have been getting away assessing acquisition of skills looking only to the end result. The artefact. Not any more. We now have to do the heavy lifting and look at the process that led to the artefact. We are coming out of the era when the artefact was a valid proxy. Not any more.
Another colleague, Shane Dawson, suggests that we make a shift to “process collection”. If the artefact is no longer a suitable proxy for knowledge and skills, then the process needs to be the focus.
What is the process of thinking with AI? First, AI serves as a type of “intellectual grunt labour”. Take DALLE-2. It can in seconds do what an artist would take hours, if not days. The artist has an idea in mind when creating. The physical (or digital) act of creation is grunt work. The conceptualizing and adjusting and tweaking is the art. In the emerging landscape of Generative AI, acts of creation sit at points where directions are provided to AI: the art prompt, the keyword edits, the revisions. To retain the value of the college essay as a form, focus needs to be on assessing and evaluating the interaction between human and artificial cognition. How do we work with, direct, influence, revise, and restructure what can be created by AI in seconds? The current process of AI ascension is similar to the advent of tools for farming. Planning, finding the right soil, planting, weeding, watering, fertilizing, harvesting – all these core tasks are the same whether using tools or not. The tools reduce grunt labour, resulting in better and bigger crops per each human’s effort. AI does something similar. AI still requires the human for input and revision. AI handles the grunt work of creation, however, faster than we could ever have imagined in many writing and art tasks. Collecting markers of knowledge during the process still enables evaluation of knowledge. This should be the primary point of assessment. Over the next decade, we’ll be going through this negotiation process in many areas of our lives: what should humans do? What should machines do? In few areas is that reckoning more urgent than in the content of college curriculum and the ways in which we assess competence and knowledge.