About
Assessment should show what people can do.
EdSphere turns written work into practical tasks that students can evidence. We want that to be the normal way to assess, not the exception. Here is why, and what the research says.
Our vision
Proof of what students can do.
Every grade a student earns should point to something they actually did: captured as they did it, explained in their own words, and checked by a qualified person.
We want setting a practical task to be as easy as setting an essay is today, in every subject, from business and law to engineering and medicine.
What we believe
Five beliefs, and the evidence for them.
Each one shapes what we build. Each source opens the original.
1
Written work no longer proves learning.
MIT’s 2026 report found that AI can now give credible answers to almost any written assignment in its undergraduate courses. Most students use AI every week, and most faculty see more cheating. An essay can’t show who did the thinking.
Source:MIT, 2026 (opens in a new tab)Lumina Foundation and Gallup, 2026 (opens in a new tab)AAC&U and Elon University, 2026 (opens in a new tab)
2
Catching AI is not the answer.
AI detectors give false alarms, and universities have switched them off. Australia’s higher education regulator, TEQSA, says institutions should redesign assessment rather than try to detect AI.
Source:TEQSA (Australia), 2025 (opens in a new tab)Vanderbilt University, 2023 (opens in a new tab)
3
Practical work shows what students can do.
MIT recommends oral exams and assignments paired with conversations. TEQSA asks for several kinds of evidence and for assessing how students learn, not only what they hand in. A Berkeley study of 95,000 students recommends practical and oral exams.
Source:MIT, 2026 (opens in a new tab)TEQSA (Australia), 2023 (opens in a new tab)UC Berkeley, Center for Studies in Higher Education, 2026 (opens in a new tab)
4
Doing the thinking is what builds it.
In an early MIT Media Lab study, people who wrote essays with ChatGPT showed the weakest brain engagement and struggled to quote their own essays. Practical tasks keep students doing the thinking, which is how they learn.
5
Practical assessment has to be quick to run.
Watching every student, asking follow-up questions, and writing it all down can take close to an hour per student. EdSphere handles the organising, so instructors spend their time on the judgement.
Research
The research we draw on.
Universities, regulators, and researchers are reaching the same conclusion: assess what students do, not only what they write.
- MITAugust 2026Report of the Ad Hoc Committee on AI Use in Teaching, Learning, and Research TrainingAI can now give “reasonable responses to almost any written assignment” in MIT’s undergraduate curriculum. The committee recommends oral exams, portfolios, and assignments paired with in-class conversations.Read the source (opens in a new tab)
- MIT Media LabJune 2025 (preprint)Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing TaskPeople who wrote essays with ChatGPT showed the weakest brain engagement of three groups and struggled to quote their own essays. A small, early study of 54 people.Read the source (opens in a new tab)
- TEQSA (Australia)September 2025Enacting assessment reform in a time of artificial intelligenceDetecting AI is not a viable answer. Institutions should redesign assessment, with secure tasks such as oral exams and in-class demonstrations.Read the source (opens in a new tab)
- TEQSA (Australia)November 2023Assessment reform for the age of artificial intelligenceTrustworthy judgements about learning need “multiple, inclusive and contextualised approaches to assessment” that emphasise the process of learning, not only the final product.Read the source (opens in a new tab)
- TEQSA (Australia)June 2026Assuring quality learning in a gen AI-integrated future: The role of adaptive capabilitiesAssure what students have learned while building the capabilities they need, such as evaluative judgement, critical thinking, and ethical reasoning.Read the source (opens in a new tab)
- UC Berkeley, Center for Studies in Higher EducationMay 2026Study of more than 95,000 students at 20 US public research universitiesRedesign assessment rather than restrict AI: in-class, oral, or practical exams, and assignments where students document their reasoning.Read the source (opens in a new tab)
- Lumina Foundation and Gallup20262026 State of Higher Education study57% of US college students use AI in their coursework at least once a week.Read the source (opens in a new tab)
- AAC&U and Elon UniversityJanuary 2026National faculty survey on generative AI78% of faculty say cheating on their campus has increased since generative AI arrived.Read the source (opens in a new tab)
- Vanderbilt UniversityAugust 2023Guidance on AI detection and why we’re disabling Turnitin’s AI detectorOne of the first universities to switch off an AI-writing detector, citing false positives.Read the source (opens in a new tab)
How we build
Six promises.
The rules we hold ourselves to.
People make the decisions.
The AI drafts and asks questions. A qualified person decides every grade, and the software makes sure of it.
Replace work, don’t add it.
The practical task replaces the written assignment. It doesn’t come on top of it.
Fit into a busy week.
Setting up a task means editing a proposal. Marking a student is one page with two choices per criterion.
Meet students where they are.
A phone browser and a private link. No app, no account, no IT ticket.
Check our own claims.
The times on this site are estimates. Our pilots measure the real numbers against each instructor’s current way of marking.
Protect student work.
Private storage, location removed from photos, and student work used only for the assessment it was captured for.
Where we’re heading
Start small, then grow.
Each step has to work before we take the next.
01
One course.
Start with one practical assignment and one instructor. Measure two things: how long it takes, and whether a second marker agrees with the grades.
02
A whole program.
Show evidence for every outcome a program promises, across all its courses, ready for review and accreditation.
03
Beyond the university.
Records of what people can actually do, useful to the employers and professions they go on to. A long way off, but we design for it now.
Try it on one assignment.
If you teach or assess and think assessment needs to change, we’d like to hear from you. Pilots are free.