Kayeh's patented AI measures both the convergence and the divergence in any body of text, from a comment thread to a Security Council transcript. It returns the principles both sides already hold, a consensus likelihood score, and bridging replies written to defuse rather than inflame.
US Patent US-11995408-B2 · Built for community, enterprise and diplomacy
Enterprise teams spend $10,000 to $150,000 a year on social listening and community platforms that classify sentiment and flag toxicity. None of that spend answers the question leadership actually asks: is this community converging on something, or coming apart? The evidence of the gap shows up in the field's own data. Proving ROI is the most cited challenge among community professionals, and misinformation and social conflict top the moderation challenge lists, in part because no existing tool measures either discourse health or manufactured division.
Disinformation alone was costing investors and consumers an estimated $78 billion a year as far back as 2019 (CHEQ & University of Baltimore), and the conversations it targets have only grown larger since.
Of community professionals cite proving ROI as their top challenge. CMX Community Industry Report, 2025.
Typical enterprise spend on listening tools that measure sentiment, not discourse health. Brandwatch pricing benchmarks.
∂The measurement gap
Sentiment analysis and social listening platforms cluster people by what they already agree on. That shows you the silos. It cannot tell you where a bridge between silos would hold. Kayeh scores every claim in a conversation along two independent dimensions, then maps the whole discussion onto a Thought Space where genuine common ground becomes visible.
How strongly a claim is echoed, reinforced and repeated across the conversation. High convergence with no divergence is conformity: agreements standardize, but overused they become authoritarian bottlenecks.
How many distinct uses and distinct evidences a claim carries. Divergence is where new ideas live, but unmeasured divergence loses focus and dilutes motivation. Kayeh is the only platform that quantifies it.
Where the two dimensions balance, Kayeh identifies principles: claims that bridge across polarized positions because both sides already use them and evidence them. Each analysis returns those principles ranked, with a consensus likelihood score.
∑How it works
A comment thread, a Slack channel, a moderated forum, a negotiation transcript, or a full Security Council proceeding. No survey design, no participant onboarding, no voting exercise required.
Kayeh's patented framework scores each claim, then extracts:
Each analysis returns a consensus likelihood score, the Thought Space map, and Socratic bridging replies you can review, edit and post to move the deliberation forward.
→A polarized thread
Both sides hold that every resident deserves reliable, affordable access to food, and both treat market structure as a means to that end rather than the end itself.
Suggested bridging replyMixed models already work at scale. Valencia's Mercado Central has run for a century with hundreds of independent stalls competing side by side inside public infrastructure. The question worth debating is which parts of that model fit here.
cWho it's for
Community managers · Directors of Community · CCOs
Proving ROI is the field's most cited challenge. Kayeh gives community teams a defensible discourse-quality metric alongside the bridging content that improves it, so the health of a community becomes something you report on, not something you assert.
T&S leads · Content policy managers
Toxicity classifiers tell you what to remove. Kayeh tells you whether a conversation is polarizing or converging, which lets policy teams intervene upstream of the incident instead of after it.
Narrative intelligence teams · Researchers
Divergence measurement separates organic disagreement from manufactured division. Claims with high repetition but artificially narrow evidence patterns surface as anomalies, including coordinated gaslighting.
Strategic communications officers · Heads of digital advocacy
From UN dialogs to national deliberations, Kayeh analyzes full proceedings and returns the principles all parties already hold, giving mediators a documented starting point rather than an intuition.
✓Grounding
A new mathematical framework for deliberative systems and computational linguistics, granted in 2024 with further filings in prosecution. The framework cannot be replicated by attaching sentiment analysis or LLM prompts to a dashboard.
We continually validate the model on real public disputes on X, generating bridging replies to polarized threads and measuring how people respond, with no paid promotion. The replies map shared principles rather than take sides.
Founders with prior exits and decades of AI R&D across search, sentiment analysis and platforms, advised by leaders from UNESCO, the Hoover Institution, Apple, Microsoft and Meta's election integrity work.
$How Kayeh is funded
Healthier public conversation is a shared good, so we price it like one. Individuals use Kayeh at no charge. Supporters who want this infrastructure to keep running contribute what it is worth to them. Organizations with deployment, security or volume requirements work with us directly on terms scoped to their situation.
Analyze the conversations you moderate, mediate or care about. The shared principles, the consensus score and the bridging replies, at no charge.
Join the listContributions keep the analysis free for the moderators, mediators, journalists and classrooms doing this work without a budget. Kayeh is a mission-driven company, not a charity, so contributions are support rather than tax-deductible donations, and we publish how they are used.
Support the workFor platforms, enterprises, NGOs and multilateral organizations. Every engagement is scoped to your deployment: on-premise installation for media-sensitive institutions, API and agent-framework access at volume, and pilot programs with defined success criteria.
Talk to us∑Pilot program
We are onboarding a limited number of pilot partners for the 1.0 release. Book a call and bring your hardest thread, transcript or community with you. We will return the analysis: the shared principles, the score, and the bridging content.
Request a pilot