The scientific basis | Digisarion Intelligence
GEI-AI draws on published work in affective neuroscience, cross-cultural psychology and the cognitive science of trust. How the platform will be validated against expert judgement.
We did not invent the science. We are applying it, and we intend to prove it.
The frameworks GEI-AI draws on are established, peer-reviewed and public. Anyone can read them. That is deliberate: a measurement instrument built on contested or proprietary theory is difficult to defend to a regulator, and impossible to defend to a scientist.
What is ours is the application. Turning that body of work into a single measurement protocol for regulated communications, and validating it against expert judgement rather than asserting it.
The traditions: Where each dimension comes from.
Trust
The cognitive science of trust and belief updating, concerned with how people decide whether a source and a message are credible, and what causes that judgement to shift.
Emotion
Affective neuroscience, including Damasio's work on somatic markers, which established that decision-making integrates emotional and cognitive processing rather than separating them (Bechara and Damasio, 2005), and Ekman's work on emotion classification (Ekman et al., 1997).
References
Bechara, A. and Damasio, A.R. (2005) ‘The somatic marker hypothesis: A neural theory of economic decision’, Games and Economic Behavior, 52(2), pp. 336–372. Available at: https://doi.org/10.1016/j.geb.2004.06.010 (Accessed: 27 July 2026).
Ekman, P., Matsumoto, D. and Friesen, W.V., 1997. Facial expression in affective disorders. What the face reveals: Basic and applied studies of spontaneous expression using the Facial Action Coding System (FACS), 2, pp.331-342.
Culture
Cross-cultural psychology, including Hofstede's cultural dimensions (Hofstede, 2009), Gelfand's work on cultural tightness and looseness, and the GLOBE programme of House and colleagues (Aktas et al., 2016).
References
Hofstede, G., 2009. Geert Hofstede cultural dimensions.
Aktas, M., Gelfand, M.J. and Hanges, P.J. (2016) ‘Cultural Tightness–Looseness and Perceptions of Effective Leadership’, Journal of Cross-Cultural Psychology, 47(2), pp. 294–309. doi:10.1177/0022022115606802.
These are not fringe positions. They are among the most cited frameworks in their fields, and they are contested in the ordinary way that established science is contested. We build to be defensible under that scrutiny rather than to assume any one camp is right.
What is ours
The synthesis, stated plainly.
Combining these frameworks into an applied protocol that scores a regulated communication, and the logic that resolves three separate measures into one rating, is the proprietary part of this work. It is not published, and it will not be published here.
We say that openly rather than being vague about it. A company that gestures at "proprietary methodology" without saying where the line falls invites the assumption that there is nothing behind it. Naming the boundary is more credible than obscuring it.
How it will be proven
Three principles, and no numbers yet.
Ground truth comes from independent domain experts, not from us.
The reference judgements the platform is measured against are produced by qualified people outside the company. A model scored against labels created by the party that built it is measuring its own opinion, and the resulting accuracy figure is not worth publishing.
Scoring is measured against expert judgement, not user self-reports.
Asking people whether they found something trustworthy produces a different and weaker signal than asking trained assessors to evaluate the communication against a defined standard.
Validation is triangulated.
No single benchmark carries the claim. The approach combines agreement between independent experts, consistency on material the model has not seen, and performance in real-world parallel running alongside an organisation's existing process. The Ada Lovelace Institute has concluded that model evaluations alone are insufficient for assurance purposes and must be combined with real-world testing, which is the position we have built to.
Where we actually are
No validation study has been completed.
The three dimensions are constructs to be validated, not proven instruments. No accuracy figure is published on this site, and none will be until it has been measured under the approach described above.
We understand that this is not the most persuasive thing a company can say about its own technology. It is the accurate thing, and for the organisations we are building this for, a supplier who publishes an unproven number is a supplier who will do it again in a compliance report.
Research partnership
Built with people who do not work for us.
The validation approach requires independent expertise, and we are working towards a formal research partnership to provide it. We will name partners once agreements are in place and not before.