Skip to main content

Diaposi Research & Development

Three shifts defining 2026

Every one of these is presented as a direction the sector is already moving in. Each also rests on something nobody has checked. What would it take to check them?

A companion to the research brief, written for funders and programme committees weighing where learning technology is actually going.

Diaposi · Research & DevelopmentWorking paper · 26 August 2026

Stated as the sector states them, then examined.

These are the directions learning technology is moving in. Each carries an unexamined assumption, and each of those assumptions is answerable with the data a Diaposi session already produces.

That is the only reason they appear on a research page. A trend is a claim about the future, and a claim about the future is worth exactly as much as the evidence anyone is willing to gather against it.

Predictive analytics in learning

Learning platforms are moving from reactive to predictive: real-time signals flag people falling behind while there is still time to help. The learning-analytics market is projected to compound at 19.97% a year to 2035.

19.97%CAGR, 2025–2035
/100live engagement score

The assumption nobody states out loud.

The question inside it

Predictive of what, validated against what? A flag that fires with no ground truth behind it is a guess with a progress bar. Nothing here should be called predictive until it has been tested against an outcome nobody chose after the fact.

Data-driven gamification and engagement

Game mechanics in learning are getting data-driven: difficulty and rewards adapt to each learner, keeping people challenged but not overwhelmed.

7poll and quiz types
3difficulties per chapter

Engagement was only ever a proxy for something else.

The question inside it

Does adaptive difficulty raise learning, or raise time-on-tool? Engagement was only ever a proxy, and a tool optimising its own proxy is the oldest failure mode in this field.

Accessibility, inclusion and ethical design

Accessibility is shifting from a compliance checkbox to a design-first concern: learning technology has to serve everyone, regardless of ability, background or context.

0accounts to participate
9interface languages

Two different claims, and only one of them is about inclusion.

The question inside it

Does removing the account barrier change who participates, or only how many? Those are different claims and only one of them is about inclusion. That is study S5.

What would settle any of them

Three trends, three testable claims, one instrument that already records what they need.

None of the three is settled by a better dashboard. Each is settled by a study, and the three studies are already scoped in the open programme: predictive claims need a ground truth nobody chose after the fact, adaptive difficulty needs an outcome measure that is not time-on-tool, and inclusion needs a comparison of who participates rather than how many.

A flag that fires with no ground truth behind it is a guess with a progress bar.

(1)

Research correspondence

The programme is open to partners.

Bring a research question, a cohort, or one lecture. We will bring the instrument, raw export and an honest account of its limits.

Discuss a research partnership