EduCreds Trust Agent

Inspiration

Educational credential fraud remains a major challenge globally. While blockchain can make certificates tamper-proof, it does not answer a more fundamental question:

Who should be trusted to issue those credentials in the first place?

As we built EduCreds, we realized that trust must be established before credentials are issued. This inspired us to create the EduCreds Trust Agent (ETA), an AI-powered governance agent designed to evaluate institutional credibility before an institution is allowed to participate in the EduCreds ecosystem.

Our goal was to move beyond simple chatbots and build an agent that can reason, evaluate evidence, and support governance decisions that impact real educational institutions.


What It Does

The EduCreds Trust Agent evaluates institutions through a framework we developed called Proof of Institutional Credibility (PoIC).

When an institution applies to join EduCreds, the Trust Agent analyzes multiple credibility dimensions including:

  • Institutional identity
  • Domain ownership and digital presence
  • Accreditation and legitimacy indicators
  • Governance and operational signals
  • Behavioral and risk factors

The agent computes a PoIC score and generates governance recommendations such as:

  • Approve
  • Approve with Monitoring
  • Restrict Issuance
  • Suspend

The resulting recommendation is then used by the EduCreds governance layer to determine whether an institution should be allowed to issue credentials and under what conditions.


How We Built It

Our architecture combines deterministic trust computation with AI reasoning.

Core Components

  • EduCreds Trust Agent (ETA) for institutional analysis
  • Gemini as the governance reasoning engine
  • Google Cloud Agent technologies for agent development and orchestration
  • Arize Phoenix for observability, evaluation, and agent introspection
  • Blockchain infrastructure for immutable governance records and credential trust

Workflow

  1. Institution submits onboarding information.
  2. ETA computes the PoIC score.
  3. Gemini analyzes the trust context and governance implications.
  4. Governance recommendations are generated.
  5. Decisions and audit records can be stored immutably for accountability and transparency.

This creates a transparent and explainable trust layer for educational credential ecosystems.


Challenges We Faced

One of the biggest challenges was understanding how to combine deterministic scoring with AI reasoning in a way that remained explainable.

We also explored multiple approaches for agent orchestration, observability, and MCP integration. Learning how to connect governance reasoning with tracing and evaluation tools required a significant architectural redesign.

Another challenge was ensuring that AI recommendations remained governance-focused and auditable rather than functioning as a generic chatbot.


What We Learned

Through this project, we learned that effective AI agents require much more than language generation.

Real-world governance systems need:

  • Explainability
  • Observability
  • Evaluation
  • Accountability
  • Human oversight

We also learned how observability platforms such as Arize Phoenix can help make AI reasoning transparent and inspectable, which is critical when AI contributes to institutional trust decisions.


What's Next

Our vision is to evolve the EduCreds Trust Agent into a comprehensive AI governance layer for academic infrastructure.

Future Plans

  • Enhanced institutional verification pipelines
  • Continuous credibility monitoring
  • Governance proposal evaluation
  • DAO-assisted trust management
  • Cross-border academic trust networks
  • Self-improving governance agents using observability and evaluation feedback

By combining AI governance, institutional trust scoring, and transparent decision-making, we aim to create a future where educational credentials can be trusted globally.

Share this project:

Updates