Understanding what people instinctively feel

The problem

Consumer-facing organisations need to know what people instinctively feel about their brands, values, and interactions. Existing methods do not reliably deliver this.

Problem 1
Survey bias

Traditional surveys ask people to declare their opinions. People answer as they wish to be seen, not as they instinctively feel. This is known as social desirability bias. No amount of questionnaire design eliminates it — it is structural.

Problem 2
Implicit tests don't scale

Implicit Association Tests partially address this by measuring reaction time as a proxy for instinctive association. But IATs are designed for binary comparisons, require laboratory conditions, and cannot be deployed at population scale.

Problem 3
Inferential Echo-loops

Current-generation LLM tools rely on historically-ingested pattern recognition rather than direct interaction with users. Petalyx operates as a proactive probe of live human sentiment, designed to reduce deliberate (System 2) reasoning through fast forced-choice responses.

The solution

Cognitif Group Limited has developed Semantic Bathymetry™: an anonymised-by-design data architecture that captures latent associative orientation at population scale, without recording personally-identifiable information.

Features
  • Petalyx™ presents users with rapid word-association tasks. For each prompt word, users choose from five response words arranged radially, plus ‘None of the Above’.
  • Both the chosen response and all non-chosen options are treated as structurally informative.
  • Petalyx™ and HMX capture latent human sentiment via intuitive low-friction GUI interfaces and remove noise and semantic bias before abstracting data as a series of numerical coordinates.
  • All individual session data is destroyed on the user's device before any transmission. Only aggregate coordinate values are transmitted and stored.
Benefits
  • Measures genuine instinctive orientation, not declared or curated opinion.
  • Captures both positive affect and systematic aversion.
  • Deployable at population scale without laboratory conditions, special equipment, or trained administrators.
  • Signal-to-noise ratio designed to outperform raw reaction-time data and self-report instruments.
  • The stored dataset is designed to contain no personal data within the meaning of UK GDPR Recital 26, with no individual-level record transmitted or retained.