Tools / Credential research
Credentials anyone can earn
The goal: anyone who has learned something, wherever and however they learned it, can have it assessed against a published standard and recognised, without first paying for a course, a seat or a degree.
Bentley Moon calls the aim the democratization of knowledge proof. Kosplora would never issue credentials itself. This page sets out how it could work, what already exists, and what would have to be true first.
How it could look
One person, one piece of work, on a route that exists today. Solid is available, dotted is available in part, hatched is not built.
Learn anywhere
A free course, the library, a job, a book, a route. Where and how you learned is not the question.
Available
Finish a piece of work
For example, rebuild Epoch AI’s growth rate for training compute on the compute route and compare your figures with the published ones.
Available
Have it assessed against a published standard
A reviewer checks the report against ACM’s Results Reproduced criteria and writes down the reasons.
The standard is named; no reviewer yet
Receive an open, signed record
A partner issues a record in an open format naming the standard, the evidence and the reviewer, which anyone can check without asking the issuer.
Not built
Keep it and decide who sees it
You hold the record. A school or employer checks the signature and decides what it counts for.
Not built
Open formats for steps 4 and 5 already exist: the W3C Verifiable Credentials Data Model 2.0, a W3C Recommendation since May 2025; Open Badges 3.0, May 2024, in which each badge is also a verifiable credential; and the Comprehensive Learner Record 2.0, March 2025, for a whole record of learning. What does not exist yet is the reviewer and the partner.
Why it matters
Learning has become cheap to find. Having it recognised has not, and removing a degree requirement does little when an employer has no other way to judge finished work.
- $11,950 and $45,000
- Average published tuition and fees for 2025 to 2026 at a public four-year college in-state and at a private non-profit one, before living costs. College Board, Trends in College Pricing 2025
- Fewer than 1 in 700
- Hires in 2023 that went to people without a degree because an employer dropped the requirement. About 45% of firms that dropped it changed their hiring in name only. Burning Glass Institute and Harvard Business School, 2024
What already works, in parts
- The Open University Open admission, with no formal entry requirements for most undergraduate courses, since its first students in 1971.
- CLEP Credit by examination for $97 an exam plus a test-centre fee, accepted at more than 3,000 colleges.
- University of the People No tuition; an accredited degree for which students pay $180 per undergraduate course assessment.
- VAE, France A national qualification awarded for what a person has learned through experience, with no minimum length of experience.
- EU recommendation on validation, 2012 Member states should let people gain a full or partial qualification from learning done outside formal education.
Each part exists somewhere: open admission, assessment without required teaching, recognition of experience, and open signed records. Kosplora’s aim is to connect them around work a person can show, checked against standards anyone can read.
From learning to recognition
Evidence, assessment, issuance and acceptance are separate responsibilities.
Evidence
The learner
Shows work, explains choices and records the help used.
Assessment
Qualified reviewers
Compare demonstrations with published criteria and explain the judgment.
Credential
A credential partner
Takes responsibility for issuing a specific claim and keeping its status current.
Recognition
The receiving institution
Decides whether to accept that claim for a stated purpose.
Acceptance is a separate decision. A credential would not automatically confer admission, credit or employment.
A course completion record, a practical demonstration and a qualification can support different claims. The proposed assessment would name the capability and its intended use before choosing tasks or a scoring method.
Practical work, explanation and repeated demonstrations could supplement conventional tests. Claims of better accuracy, fairness or predictive value than ACT- or SAT-type tests would need comparative evidence. No such result is claimed here.
Evidence of understanding
Combine observations that answer different questions about a specific capability.
Work and revision
What can the person produce?
A design, investigation, performance or working solution, with sources, intermediate work and revisions.
Limit: A finished artifact alone does not establish authorship or understanding.
Explanation
Can they account for the work?
Explain methods, defend choices, identify mistakes and describe contributions from tools or other people.
Limit: A rehearsed explanation may not carry over to a different problem.
An unfamiliar task
Does the understanding transfer?
Apply the ideas to a new case under declared conditions, with agreed tools and accessibility support.
Limit: One performance supports a limited claim about capability.
A later demonstration
Does the capability persist?
Demonstrate it on another occasion or in another setting, using criteria appropriate to that capability.
Limit: The record must remain specific about the settings and time observed.
Independent understanding and effective use of tools
Both can be worth demonstrating. An assessment of independent reasoning would state which sources and tools are allowed. An assessment of work with AI could examine how a person directs the tool, checks its output, catches errors and takes responsibility for the result.
Those are different capability claims. Permitted assistance, collaboration and accessibility arrangements must be declared before the assessment and recorded with the result.
Illustrative assessment design
Investigating a public claim
Can a person trace a claim to evidence, assess the sources and explain uncertainty?
This is a worked design example. It is not an available assessment or an awarded qualification.
It connects Kosplora’s digital-agency work with the source checking used in Values Commons.
- Prepare
- Investigate a public claim and keep a source trail. Record help from search tools, collaborators and AI.
- Explain
- Discuss the evidence and a plausible alternative explanation with a reviewer. Outside coaching is not permitted in this part.
- Apply
- Examine a different claim using a supplied evidence set, under the tools and accessibility arrangements agreed in advance.
- Revisit
- Demonstrate the capability later. Reviewers record reasons against the criteria and retain disagreements for review.
Proposed review criteria
- Source tracing
- Find the original claim and underlying evidence; identify author, date and later corrections.
- Evidence quality
- Distinguish observations, estimates and interpretations; explain what the methods can establish.
- Alternatives
- Consider credible conflicting evidence and explain what would change the conclusion.
- Uncertainty
- State a conclusion no stronger than the evidence, including what remains unresolved.
This outline still needs task development, reviewer calibration and evaluation. No pass threshold has been established. Agreement with the Institute’s views would not be an assessment criterion.
What a credential should contain
A specific claim with enough context for someone else to judge its relevance.
A signed credential can help establish who issued a record and whether it has changed. The evidence behind it, the quality of the judgment and the recipient’s decision to accept it require separate scrutiny.
- Capability and purpose
- The specific claim, intended use and published criteria, including their version.
- Evidence and conditions
- Relevant work and observations; task versions, dates, permitted assistance and necessary assessment conditions.
- Judgment and responsibility
- The result, reasons, reviewer qualifications, conflicts of interest and issuing organization.
- Status and corrections
- How to check authenticity and current status; any expiry or review date, correction history and withdrawal rules.
- Sharing and appeal
- What the learner can export or disclose, how evidence is protected and how to challenge or correct a decision.
Proposed information requirements for a future credential.
Relevant standards: Open Badges 3.0 and the W3C credential model.
Identity and unauthorized assistance
Integrity checks need evidence about both missed assistance and false accusations.
Research must consider impersonation, remote coaching, concealed in-ear devices and undisclosed AI assistance. No validated device-detection system is available from Kosplora.
Set the conditions in advance
Name the capability, permitted tools, identity checks and observation conditions. Explain what is collected, who can see it and when it is deleted.
Evaluate the checks
Compare supervised demonstrations, live follow-up questions and detection methods. Measure missed cases, false flags, access barriers and privacy costs for the actual setting.
Review a concern
An automated flag would prompt review. It would not by itself establish misconduct or absence of competence. The person should be able to explain the circumstances and challenge the evidence.
Provide another route
Accommodate hearing aids and other assistive tools. Where a method is unsuitable, provide an equivalent assessment arrangement with published criteria.
No single check can establish that all unauthorized assistance was absent. The record should state the conditions and remaining uncertainty.
Access, control and recognition
People should be able to seek recognition for learning without first buying a particular course or attending a prescribed institution.
Learner control
- Public criteria and examples before an assessment.
- Clear costs for assessment, review and reassessment.
- Accessible formats and agreed assistive tools.
- Exportable records and control over evidence shared.
- Reasons for a decision, corrections and an appeal route.
Accountable recognition
A credential partner would need to specify what it assesses, who may assess it, how conflicts of interest are handled and how records remain usable if the issuer changes or closes.
A school, employer or collaborator would decide what it accepts and for what purpose. File compatibility does not establish academic credit, accreditation, licensing or hiring eligibility. No recognition agreement is announced.
Learning and creative exploration should remain worthwhile without assessment. The longer-term aim is to help people demonstrate capabilities, teach one another and contribute to shared work. Kosplora does not propose a universal score of intelligence or personal worth.
Research before recognition
The next substantive work is to define and evaluate an assessment for a specific capability and intended use.
- Validity
- Does the evidence support this capability claim for this use and population?
- Consistency
- Do reviewers and repeat demonstrations reach sufficiently consistent judgments?
- Fair access
- Do cost, language, disability, connectivity or assessment format create unrelated barriers?
- Integrity
- How often do checks miss unauthorized assistance or wrongly flag legitimate work?
- Useful recognition
- Who would accept the result, for what purpose, and what would assessment and reassessment cost?
Published criteria, task examples and reviewer guidance would be prepared before any assessment service opens. Studies would need suitable expertise, consent, fair participation conditions and an analysis plan. Simulated learners or software tests cannot establish the educational validity of an assessment.
Standards informing the proposal
- Standards for Educational and Psychological Testing
AERA, APA and NCME · 2014
Validity for a stated interpretation and use, consistency and fairness. A basis for assessment research; it does not validate this proposal.
- Open Badges 3.0
1EdTech
A standard for portable achievement records, including criteria and evidence. Format support does not establish assessment quality or acceptance.
- Verifiable Credentials Data Model 2.0
W3C · 2025
A model for verifiable claims and privacy considerations. Verifying a record’s authenticity does not establish the truth of its claims.
- Credential Transparency Description Language
Credential Engine
Shared descriptions of credentials, competencies, assessments and their relationships. Useful for making standards and requirements inspectable.
Proposal updated .