Scaffold or Crutch? Framing AI through Expectancy-Value Theory

This is the second post in a brief series that contextualizes AI in education within prominent theories and phenomena of educational psychology. The first post discussed AI and flow theory. Today we’ll focus on Expectancy-Value Theory.

Expectancy-Value Theory (EVT) describes a learner’s motivation on a given academic task as a function of their expectations of success on the task as well as of their perceived value of the task. The basic premise is that these components combine multiplicatively to yield motivation, i.e.

Motivation = Expectancy x Value

What follows is that if a learner expects that they can succeed on a given task and they perceive that task as valuable, they will be highly motivated to pursue it. Conversely, if either expectations of success or value are low, this will yield lower motivation. And if both are low, then our hypothetical student probably won’t be very motivated to pursue the given task at all.

N.b. If you’re interested in reading more on EVT, check out basically anything from Eccles & Wigfield. Their article from 2000 in Contemporary Educational Psychology is probably the definitive citation, though.

Flavors of Value

According to EVT, there are different flavors of task value. These include intrinsic value, utility value, and attainment value.

Intrinsic value represents a learner’s innate interest in the task. For example, if you’ve got a kid who likes baseball, and you assign him to read a novel about baseball, this will have high intrinsic value to him.

Utility value represents a learner’s perception that accomplishing the task at hand will help them accomplish some other important task in the future. For instance, I might ascribe high utility value to completing my Algebra 2 homework if I think that mastering Algebra 2 will be necessary for me to get into the engineering program at my dream college.

Attainment value is a little bit less straightforward. It represents the extent to which a task aligns with or reinforces a central part of their identity. A student who considers herself to be “a writer” will perceive writing tasks as more valuable. Or a student who sees herself as a “math person” will feel the same way about math tasks.

EVT also includes cost as a sort of “negative value.” Cost refers to something that a learner must “spend” to accomplish the task, including time, effort, or emotional energy. This can also encompass opportunity cost, where performing a task necessarily means that the person loses the opportunity to do something else (e.g. the opportunity cost of studying might be foregoing a trip to the movies with friends). Typically, it's better for a task to have lower costs, although any worthwhile task must have some cost associated with it to foster motivation. If a task is truly costless, then it’s hard to feel a sense of ownership.

Note: I finished by PhD in 2019, and I’m basically not up-to-date on the literature since then. I think there’s been some movement in how cost is conceptualized over the past 5 or 6 years that isn’t reflected here. Nevertheless, the above is probably close enough for an informal blog.

EVT and AI

By framing student motivation using EVT, we can hypothesize how AI tools might influence students’ motivation. There seem to be strong arguments for how AI could promote motivation as well as for how it could hinder motivation when considered through this lens.

Let’s start with the positives.

Probably the most obvious way that AI can foster students’ motivation is by improving their expectations of task success. If a student is struggling to complete a math homework assignment, for instance, he might have low expectations of success. Rather than giving up on the assignment, he could turn to an AI tool for help on the task. If used appropriately (i.e. as a scaffold), the AI tool might correct misunderstandings and provide the sort of feedback that a tutor would. This support could change the student’s expectations for success by showing him that he can, in fact, complete the assignment, which would lead to improved motivation.

This, of course, assumes the student uses the AI tool as a tutor rather than as an “answer engine.” The best case here seems to be targeted use intended to get over a particular sticking point.

Another way in which AI tools can increase motivation is by decreasing the perceived cost of performing the given task. AI tools can reduce the time, effort, and emotional cost associated with completing a given task – particularly a task that may be slightly too difficult for a student – thereby increasing the student’s motivation to begin and persist throughout the task. For example, if I suspect I’m going to get frustrated while trying to do my Algebra homework (i.e. I’ll incur some emotional cost), I’m less likely to be motivated to do it. But if I know I can preempt my frustration by turning to my friend Gemini, then I might anticipate less emotional cost.

Let’s move onto the negatives. What’s interesting here is that the first two negatives are basically the two positives above taken to their extremes.

Although targeted use of AI might help students get over sticking points and increase expectations of task success, prolonged use probably does the opposite. Over-reliance on AI to complete assignments will slowly erode expectations of success over time. That is, students won’t judge themselves capable of doing the work without AI if they use it too much.

Second, if a task truly becomes “costless,” students can’t really feel motivated to complete it. Imagine the case where a teacher assigns a worksheet of math problems. If a student can simply take a picture of the worksheet on their phone, upload the picture to Gemini, and write a 1-sentence prompt asking Gemini to solve the problems, at what point in this whole endeavor will the student feel motivated?

A related issue is that the presence of AI tools will likely decrease the utility value of many tasks for many students. Let’s tweak our previous utility value example. If I think learning how to find the derivative of a function will be an important skill for me to master if I want to be an engineer, then I’ve ascribed utility value to my calculus homework. But if I think that I’ll always just be able to ask AI to find derivatives for me, then I’ll ascribe much less value to my calculus homework.

All in all, I suppose the takeaway here is that AI tools are probably only beneficial for student motivation if used in targeted, discrete ways. Otherwise, they seem likely to hinder long-term motivation by decreasing long-term expectations of success, decreasing the perceived utility of various academic skills, and undermining ownership by making learning tasks “costless.”

If you’re enjoying reading these posts, please consider subscribing to the newsletter by entering your email in the box below. It’s free, and you’ll get new posts to your email whenever they're published.