Useful principles. Visible limits.
Cognism combines research, practitioner frameworks and classical ideas. Inclusion is not proof that a principle works in every situation. Evidence labels describe the kind of support, not a guarantee of quality or fit.
Meta-analysis
A synthesis of many studies supports the core claim.
Experimental
Controlled experiments support the core claim.
Observational
Field, longitudinal, diary or survey research supports the core claim.
Practitioner
An expert framework from practice: plausible and widely used, not formally tested.
Classical
Classical philosophy or rhetoric: enduring, not empirically tested.
Library first
Connected agents receive a policy to use relevant library principles first, cite their code and name, disclose evidence strength, and include cautions. An attributed source without a linked reference is not an independently verified academic citation.
When the library falls short
For uncovered parts of a question, the policy permits only approved scholarly catalogue records returned by the MCP server. Agents must keep this fallback evidence separate from library principles, cite the supplied source and explain its limits. An academic-looking link found elsewhere does not qualify. If no approved record covers the gap, the agent should withhold that claim or request a source review.
What a connection can do
The MCP server provides principles and citation instructions. It cannot control every answer from another AI app, verify everything that app finds, or guarantee that the app follows the policy. Check important claims before acting.