Most educational technology is built around engagement metrics: logins, completions, time-on-platform. These are easy to measure and easy to report. They are also a poor proxy for learning.
Potentially is built around a different question: what does the research say actually produces capability? That question changes what you build. It changes which features matter, which don’t, and why. It changes where you put the friction and where you remove it. It changes what you count and what you don’t.
The design choices in Potentially are not opinions. They are derived from decades of learning science. This piece sets out seven bodies of research that shaped the platform most directly — and explains what each one means in practice.
Self-efficacy and the power of evidence.
The most reliable source of confidence is not encouragement. It is evidence of having done something.
Albert Bandura identified four sources of self-efficacy — a person’s belief in their capacity to achieve a particular outcome. The strongest, by a significant margin, is mastery experience: the accumulated record of having actually done something and succeeded. Verbal encouragement from others is weaker. Watching others succeed (vicarious learning) is weaker still. Physiological state — how anxious or calm you feel — is weakest of all.
This has a direct implication for how a portfolio should work. A record that simply lists activities or claims skills provides no mastery experience. It produces a document, not a capability. The student who writes “good communicator” on their CV has not experienced the mastery loop. The student who can point to a presentation they delivered, the feedback they received, and the reflection they wrote about what changed as a result — has.
The distinction is not semantic. It is the mechanism by which a record builds genuine capability rather than documented activity.
Depth over volume.
Bloom, 1956 · Anderson & Krathwohl, 2001
Surface-level engagement and deep cognitive engagement are not the same thing. Volume does not imply depth.
Bloom’s taxonomy distinguishes six cognitive levels — from remembering at the base, through understanding, applying, analysing, evaluating, to creating at the apex. Most educational technology rewards activity at the lower levels: logging events, completing checkboxes, submitting artefacts. This conflates volume with learning.
A student who submits thirty entries that describe what they did has done something categorically different from a student who submits eight entries that demonstrate they can analyse an experience, evaluate their own performance against a standard, and articulate what they would do differently. The first student has produced more data. The second student has learned more.
Anderson and Krathwohl’s 2001 revision made an additional distinction that matters here: the difference between cognitive process — what mental operation is being performed — and knowledge type, what kind of knowledge is being applied. Metacognitive knowledge, knowing how you learn and recognising your own strengths and gaps, sits at the top of the knowledge dimension. It is the hardest to develop and the most predictive of long-term capability growth.
The reflective cycle.
An experience that isn’t reflected on doesn’t become learning. It becomes history.
Kolb’s experiential learning cycle has four stages: concrete experience, reflective observation, abstract conceptualisation, and active experimentation. The cycle only closes if all four stages occur. Most portfolio tools skip the reflection stage entirely — students log events and submit artefacts without being required to articulate what the experience meant, what changed because of it, or how they would approach a similar situation differently.
Without the reflective observation stage, the cycle doesn’t close. The concrete experience remains a memory rather than becoming a learning event. The student has done something but has not yet learned from it in any transferable sense.
This is not a minor omission. Reflection is not a nice-to-have layer on top of the portfolio. It is the mechanism by which experience becomes capability. Remove it and you have a filing system, not a development tool.
Kolb’s cycle also has implications for how reflection should be structured. Abstract conceptualisation — the stage where the student connects the experience to broader frameworks, principles, or patterns — requires some scaffolding. Students who are asked simply to reflect often produce description rather than analysis. The prompt matters as much as the reflection itself.
AI and metacognitive laziness.
Fan et al., 2025 · Risko & Gilbert, 2016
AI that reduces cognitive effort in the short term reduces learning outcomes in the long term.
This is now one of the most important findings in educational technology research, and it is consistently misunderstood. The issue is not that AI produces bad outputs. Often it produces better outputs than the student would have produced unaided. The issue is what happens to the student’s capability development when the hard cognitive work is offloaded.
The mechanism is well documented in the metacognition literature and has been confirmed by recent empirical work on AI-assisted learning. When a student uses AI to generate a reflection, draft an analysis, or articulate a skill, two things happen at once: the output improves and the learning degrades. The struggle that would have produced capability doesn’t occur. The student gets a better document and develops less.
The mechanism has a name in cognitive science. Risko and Gilbert call it cognitive offloading: using a tool or an action to reduce the mental work of a task. Their review of the evidence describes a consistent trade — offloading improves performance on the task in hand and reduces what is retained from it. A calculator, a written note and a language model all sit on the same line; they differ in how much of the thinking they take.
Fan and colleagues describe the educational form of this as metacognitive laziness — a pattern in which AI assistance reduces the metacognitive effort that produces durable learning. The student who uses AI to write their reflection has produced a better reflection and learned less from the experience than the student who wrote it themselves, even if the unaided version was rougher.
This is not an argument against AI in education. It is a technical description of where AI helps and where it harms — and that distinction should drive every design decision in a platform that uses it.
Feedback that answers three questions.
Feedback is one of the strongest influences on learning, and one of the most variable. What it says matters more than that it is given.
Hattie and Timperley’s review of the feedback literature set out a model that has held up. Feedback is effective when it answers three questions for the learner: where am I going, how am I going, and where to next. And it works at four levels — the task, the process behind the task, the learner’s own self-regulation, and the learner as a person.
The levels are not equal. Feedback about the person — praise, in most classrooms — carries the least information about what to do differently and has the weakest effect on learning. Feedback about the process and about self-regulation carries the most, because it transfers: a student told how to structure an argument can use that on the next assignment, while a student told the argument was good cannot.
The third question is the one most feedback leaves unanswered. A mark says how you went. A comment says what was wrong. Neither says what to do next, and next is where the learning is.
Learning is a cycle the student runs.
The students who do best are not the ones who are told the most. They are the ones who set goals, watch themselves work, and adjust.
Barry Zimmerman’s account of self-regulated learning describes three phases that feed each other in a loop. Forethought: setting a goal and planning how to reach it. Performance: doing the work while monitoring how it is going. Self-reflection: judging the result against the goal, and changing the approach for next time. The loop is the point — each pass through self-reflection shapes the next forethought.
Two findings matter for how a platform should behave. Self-regulation is learnable: students taught to set goals, self-monitor and self-evaluate do it more, and do better. And it depends on the learner being able to see their own performance clearly. A student who cannot see where they are cannot judge the gap to where they want to be.
Most educational technology reports on students. Zimmerman’s work argues for the reverse: showing students their own record, and asking them to act on it.
The portfolio as a high-impact practice.
Kuh, 2008 · Watson, Kuh, Rhodes, Light & Chen, 2016
A portfolio done properly is not a filing cabinet. It is one of a small number of practices that reliably change what students get from university.
In 2008 George Kuh identified ten high-impact practices — first-year seminars, learning communities, writing-intensive courses, collaborative projects, undergraduate research, diversity and global learning, service learning, internships, capstone projects, and common intellectual experiences. What they share is more telling than the list: each demands sustained effort, puts students in substantive contact with staff and peers, gives frequent feedback, and asks students to apply what they learn in new settings and reflect on it.
In 2016 the Association of American Colleges and Universities added an eleventh: the ePortfolio. The argument made by Watson, Kuh and their co-authors was that the portfolio is the practice that makes the other ten visible — the place where a placement, a project and a capstone become one connected record — and that its impact depends on the reflection and integration it asks of the student, not on the software that holds it.
That last clause is the one to keep. A portfolio that collects artefacts is a storage system. A portfolio that asks the student to connect them, reflect on them and show them to someone is a high-impact practice.
Why this matters now.
Most of this research was not written with AI in mind. Bandura published self-efficacy theory in 1977. Bloom’s taxonomy is seventy years old. Kolb’s cycle predates the personal computer. They describe something more fundamental than any particular technology: the conditions under which human beings actually develop capability.
What has changed is the urgency. As AI takes on more of the cognitive work that used to constitute employment, the question of what humans can do — and how they can demonstrate it — becomes more consequential, not less. A generation of students is entering a labour market that is changing faster than any curriculum can track. The record they carry out of higher education matters more than it ever has.
Potentially is built on the conviction that the answer to that challenge is not more data. It is better evidence — evidence that is grounded in what the research says actually produces capability, verified by people who can attest to it, and owned by the student long after graduation.
That is what the platform is for.
References.
- Bandura, A. (1977). Self-efficacy: Toward a unifying theory of behavioral change. Psychological Review, 84(2), 191–215. doi:10.1037/0033-295X.84.2.191
- Bloom, B. S., Engelhart, M. D., Furst, E. J., Hill, W. H., & Krathwohl, D. R. (1956). Taxonomy of educational objectives: The classification of educational goals. Handbook I: Cognitive domain. New York: David McKay.
- Anderson, L. W., & Krathwohl, D. R. (Eds.) (2001). A taxonomy for learning, teaching, and assessing: A revision of Bloom’s taxonomy of educational objectives. New York: Longman.
- Kolb, D. A. (1984). Experiential learning: Experience as the source of learning and development. Englewood Cliffs, NJ: Prentice Hall.
- Fan, Y., Tang, L., Le, H., Shen, K., Tan, S., Zhao, Y., Shen, Y., Li, X., & Gašević, D. (2025). Beware of metacognitive laziness: Effects of generative artificial intelligence on learning motivation, processes, and performance. British Journal of Educational Technology, 56(2), 489–530. doi:10.1111/bjet.13544 · open-access preprint
- Risko, E. F., & Gilbert, S. J. (2016). Cognitive offloading. Trends in Cognitive Sciences, 20(9), 676–688. doi:10.1016/j.tics.2016.07.002
- Hattie, J., & Timperley, H. (2007). The power of feedback. Review of Educational Research, 77(1), 81–112. doi:10.3102/003465430298487
- Zimmerman, B. J. (2002). Becoming a self-regulated learner: An overview. Theory Into Practice, 41(2), 64–70. doi:10.1207/s15430421tip4102_2
- Kuh, G. D. (2008). High-impact educational practices: What they are, who has access to them, and why they matter. Washington, DC: Association of American Colleges and Universities.
- Watson, C. E., Kuh, G. D., Rhodes, T., Light, T. P., & Chen, H. L. (2016). ePortfolios – The eleventh high impact practice. International Journal of ePortfolio, 6(2), 65–69. aacu.org/ijep
For further reading on the research behind Potentially, contact us at hello@potential.ly.
