Define the problem
Who shaped the research question and priorities?
TRUST TRACE: Research Signals
How does trust research cross boundaries?
Explore four observable connections, from the fields that cite one another to the people who publish together.
The historical collection awaits relevance screening. AI membership is a provisional title screen. Affiliation indicators are proxies. Values describe different denominators and should not be averaged or ranked.
Loading the local research evidence…
01 / Follow the connections
Explore the pattern.
02 / A historical view
Explore the five years.
The same definition for each year. Select a column to inspect its records. Changes in coverage and field composition can affect the pattern.
Fixed 0–100 scale. Sparse values are withheld. This is a historical description, not evidence of improvement or a current trend.
03 / Inspect the records
The evidence behind the signal.
Search and evidence-view buttons filter this list. Only the publication period and primary-field controls change the headline denominator.
04 / Open definitions
How this signal is calculated.
How sensitive is the choice?
| Definition | Default value |
|---|
Sensitivity results use the fixed 2019–2023, all-fields view. They are different definitions, not confidence intervals.
Shared collection and missing-data rules
All analyses use the preserved historical recovery. The underlying citation universe contains 25,328 eligible works dated 1990–2023, after excluding 58 cached retraction flags and 22 nonretracted works in the incomplete 2024 slice. The four signals select citing or coauthored papers from 2019–2023. References to later-year works and self-work references are withheld; repeated reference IDs within a paper count once. No 2024–2026 discovery candidates enter these values.
At least 100 mapped links are required to display citation percentages, and at least 30 assessable papers for affiliation percentages. These are display conventions, not validated statistical thresholds. Missing evidence is not a zero. Source classification, missing affiliations, coverage and publication practices can affect all indicators.
Why these four, and what comes later?
Connections, AI research, societal coauthorship and international coauthorship make different parts of the Foundation's mission visible. Open-access status is useful supporting context, but not a connection score. Portfolio diversity measures breadth, not interaction. Current research momentum needs a maintained current collection. Research-to-practice claims and framework classifications need reviewed content. There is no combined score or pass mark.
05 / The transdisciplinary ambition
Partnerships are a signal. Co-production needs evidence.
The cross-sector number describes affiliated coauthors. A fuller transdisciplinary assessment asks what scientists and societal partners actually did together.
Integrate knowledge
How were disciplinary and societal contributions combined?
Evaluate together
Who judged the findings and their practical relevance?
Process score: not yet estimable
The existing 30-paper abstract pilot could not adequately assess any of these three phases. This is missing evidence, not evidence of zero transdisciplinary research. The pilot has one AI coder and awaits independent human review.
The TRUST framework adds Trustworthiness, Risk, User, Sphere and Terrain as content lenses. These describe what research concerns; counting more lenses does not make a paper better. Self-trust, interpersonal, institutional and societal research remain in the intended scope alongside AI and neuroscience.
Explore the Foundation's framework →06 / A transparent research programme
What would make this useful over time?
A dated edition, one well-supported research question, and the records behind it. A newsletter could follow a connection between AI research and psychology, or examine a documented partnership across sectors. Updates should distinguish new papers, better metadata and revised interpretations.
Methods informing the design
- Purnell (2025), inter-stakeholder co-publishing. DOI: 10.1007/s11192-025-05367-2. Our definition differs from its affiliation-share threshold.
- Aksnes, Karlstrøm and Piro (2026), limits of bibliometric interdisciplinarity measures. DOI: 10.1007/s11192-025-05525-6.
- Steelman and colleagues (2021), assessing transdisciplinary research practices. DOI: 10.1007/s11625-020-00901-y.
- Krueger and colleagues (2025), the Foundation's transdisciplinary trust framework. DOI: 10.1057/s41599-025-05481-9.
Experimental design. Not a validated scientific index. The full metadata load will move to a server-side data service.