Trustweave

AI-powered crisis communication platform that combines reliable evidence and AI analysis to help organizations create clear, trustworthy, and culturally relevant public messages.

Comment

Inspiration

In a world where information travels across borders in seconds, one of the biggest challenges is no longer just accessing information—it is knowing what to trust, understanding what it means, and communicating it in a way that helps people make informed decisions.

This is a documented global challenge.

The World Health Organization (WHO) describes an infodemic as an overabundance of information, including false or misleading content, that can cause confusion, harmful health decisions, and mistrust during disease outbreaks. This demonstrates why reliable information alone is not always enough; people also need clear, useful communication that helps them understand what actions to take.

Source: WHO — Infodemic

The United Nations, through its Global Principles for Information Integrity, has highlighted how misinformation, disinformation, and hate speech can undermine public health, human rights, peace, and trust. The UN also identifies transparency, public empowerment, and societal trust as important principles for a healthier information environment.

Source: United Nations — Global Principles for Information Integrity

The challenge becomes even more urgent during disasters and emergencies. The United Nations Office for Disaster Risk Reduction (UNDRR) explains that harmful information can undermine protective actions, disrupt response operations, and erode trust. Its guidance emphasizes the importance of understanding how people access, interpret, assess, and act on information during a crisis.

Source: UNDRR — Harmful Information in Disasters: Risk Communication for Safer Decisions

The WHO also emphasizes that emergency risk communication must account for people's concerns, perceptions, beliefs, and circumstances—not simply deliver facts. Effective communication helps people make informed decisions and take protective action.

Source: WHO — Emergencies: Risk Communication

These findings led us to a deeper question:

What if an intelligent system could help organizations connect reliable evidence with communication that is clearer, more relevant, and easier for different audiences to understand?

That question inspired TrustWeave.

We set out to build an intelligent crisis communication system that brings evidence assessment, AI-generated messages, cultural context, and quality review into one workflow. Rather than treating fluent AI-generated text as automatically trustworthy, we wanted to explore a more responsible approach—one that makes uncertainty visible, keeps evidence separate from cultural signals, and leaves final decisions in human hands.

Our inspiration is simple: during a crisis, a message should do more than sound convincing. It should be grounded in available evidence, understandable to the people receiving it, and reviewed before it is used.

TrustWeave aims to help make reliable information not only available, but also understandable and useful when it matters most.

What it does

TrustWeave is an intelligent crisis communication system that helps transform reliable evidence into clear, trustworthy, and context-aware public messages.

The current version focuses on an educational public-health use case based on a configured CDC source about hand hygiene.

Here is how it works:

  • Claim assessment: Users submit a claim and define their communication objective and intended audience.
  • Evidence retrieval: TrustWeave retrieves information from its configured source and assesses the claim within the limits of the available evidence.
  • Cultural context: When a cultural reference is provided, the system can explore related cultural signals to inform alternative ways of presenting a message.
  • AI-generated communication: Google Gemini helps generate a factual baseline message and a context-aware alternative.
  • Quality review: A separate AI review step examines the generated drafts against the available source, while deterministic checks identify issues that may need attention.
  • Human oversight: Users can inspect the evidence, compare drafts, review warnings, edit the content, and prepare it for human approval.

Claims outside the supported evidence scope are marked as Not yet verified rather than being presented as confirmed.

TrustWeave does not automatically publish messages. Its purpose is to support better communication while keeping evidence, uncertainty, and human judgment visible.

How we built it

We built TrustWeave as a full-stack web application using React 19 and Vite 8 for the frontend, with Node.js and Express 5 handling the server-side workflow.

We use Zod to validate incoming requests and help enforce input limits. The Express API coordinates evidence retrieval, cultural-context lookups, AI generation, and review before returning the results to the browser.

Google Gemini powers the message-generation and independent-review stages. Qloo's cultural intelligence capabilities allow the application to explore related entities and cultural signals that may inform alternative message presentations.

We deliberately separate evidence assessment from cultural context. Cultural signals can help inspire a different communication approach, but they are not treated as proof that a claim is true or as evidence of what an entire community believes.

We also use Vitest for unit tests and mocked HTTP workflow tests, allowing us to test important parts of the application without requiring live provider calls for every test.

Our approach is to make the workflow understandable and reviewable rather than treating AI-generated content as automatically correct.

Challenges we ran into

One of our biggest challenges was combining evidence assessment, cultural context, AI generation, and quality review without confusing their responsibilities.

A message can be culturally relevant but factually incorrect. It can also be factually accurate but difficult for its intended audience to understand. We needed to keep these concerns separate while making them work together in one workflow.

Another challenge was handling uncertainty responsibly. We did not want the system to claim that it had verified information when the configured source could not support that conclusion. This led us to keep the current evidence scope deliberately limited and label unsupported claims as Not yet verified.

Integrating external APIs also introduced practical challenges involving credentials, provider availability, request validation, and testing. We addressed these through server-side integrations, input validation, and automated tests using mocked HTTP workflows.

We also had to think carefully about human oversight. In crisis communication, a convincing message is not necessarily a safe or reliable one. The system therefore supports review and editing instead of automatically approving or publishing its output.

Accomplishments that we're proud of

We are proud to have built a working foundation that brings several important stages of crisis communication into one application.

Our key accomplishments include:

  • Building a full-stack application that coordinates evidence retrieval and AI-assisted communication.
  • Creating a workflow that compares a factual baseline draft with a context-aware alternative.
  • Integrating cultural intelligence as an optional communication aid rather than a substitute for factual evidence.
  • Separating AI-generated content from an additional review stage and deterministic checks.
  • Designing the current evidence workflow to acknowledge its limitations instead of presenting unsupported claims as verified.
  • Keeping provider credentials on the server side and implementing request validation.
  • Developing automated tests for important parts of the workflow.

Most importantly, we are proud of the principle behind the project: AI should help people communicate reliable information more effectively without removing the need for evidence, transparency, and human responsibility.

What we learned

Building TrustWeave taught us that creating a useful AI application involves much more than connecting a language model to a user interface.

We learned the importance of designing the complete workflow around an AI result: where the evidence comes from, what the system can reasonably conclude, how uncertainty is communicated, and how a person can review the output before using it.

We also explored the distinction between cultural intelligence and factual verification. Cultural context can help inspire alternative ways to present a message, but it cannot establish whether the underlying information is true or accurately represent every person in a community.

On the technical side, we gained experience coordinating external APIs, validating requests, structuring server-side workflows, and testing application behavior without depending entirely on live services.

Above all, we learned that responsible AI systems should make their limitations visible and support human decisions rather than encouraging people to trust automated output blindly.

What's next for TrustWeave

TrustWeave currently starts with a focused public-health communication use case. Our next goal is to develop the system into a more flexible platform that can support a wider range of communication needs.

Our planned areas of development include:

  • Broader evidence sources: Explore ways to support additional trusted sources and more communication topics.
  • Stronger evaluation: Improve how generated messages are assessed for factual support, clarity, consistency, and suitability for their intended purpose.
  • More flexible workflows: Support different communication objectives and audience contexts while preserving evidence boundaries.
  • Better human review: Make it easier for reviewers to understand why a draft was generated, which evidence supports it, and what still requires attention.
  • More robust testing: Expand testing for provider failures, unsupported claims, edge cases, and the complete communication workflow.
  • Responsible expansion: Explore additional public-interest use cases without presenting TrustWeave as a replacement for professional judgment or official emergency communication systems.

Our long-term vision is to help organizations communicate important information more clearly when it matters most.

TrustWeave is built around a simple idea: reliable information should not only be accurate—it should also be understandable and useful to the people who need it.

Built With

Share this project:

Updates

Submission history