A Confession About the Trust Scientists

On what gets measured, what gets published, and what actually gets remembered

After six years and 140 conversations about trust, I have to admit something that has been nagging at me for a while. I hope my academic friends will forgive me by the end of this piece. Some of them may even agree with me.

When I started TrustTalk in 2020, I did what any sensible beginner does: I went to the people who know. The trust scholars. The professors. The members of FINT, the First International Network on Trust, where researchers from management science, psychology, sociology and economics meet to compare notes on the one subject I had decided to build a podcast around. They were generous with their time, and they taught me the map of the territory: what trust is made of, how it grows, how it breaks, why repairing it is so much harder than building it.

I owe them a great deal. And yet, lately, I notice something when a new academic paper on trust lands in my inbox. My eyes glaze over somewhere around the words “construct validity.”[1]

How the scientists actually study trust

Let me explain what trust researchers actually do all day, because it deserves respect before it deserves teasing.

Most trust research rests on questionnaires. Not casual ones, but carefully validated survey scales, refined over decades. Respondents rate statements on a scale from one to seven: “I believe my supervisor has my best interests at heart.” The famous model behind many of these scales - developed by Roger Mayer, James Davis and David Schoorman in 1995 - breaks trustworthiness down into three ingredients: ability, benevolence and integrity. Thousands of employees, patients, citizens and students have filled in these forms, and the patterns in their answers are what much of the field is built on.

Then there are the trust games, borrowed from experimental economics. In the classic version, I hand you ten euros and you choose how much to pass to a stranger. Whatever you send is tripled on the way. The stranger then decides how much, if anything, to return to you. How much you sent in the first place is taken as the measure of your trust. Students in university labs have played this game tens of thousands of times, in every conceivable variation. One of the more amusing findings: most people cooperate - the exception being economists themselves, whose formal training in rational self-interest seems to override the ordinary human instinct to trust.[1]

Add to this vignette studies (people react to hypothetical scenarios), longitudinal panels (the same people surveyed year after year), and meta-analyses (studies of studies, pooling the findings of many papers into one overall verdict), and you have the toolkit of the trust scientist.

This is serious, careful, necessary work. Without it, we would still be exchanging folk wisdom about trust the way we exchange opinions about the weather.

Why it makes for difficult listening

Here is my problem. TrustTalk is a 30-minute audio podcast listened to by people all over the world, many of whom, like me, did not grow up speaking English. And no one has ever been moved by a confidence interval.

When a researcher explains that perceived benevolence moderated the relationship between voice behavior and affective trust, controlling for propensity, they are being precise. Precision is their job. But somewhere between the university office and your earbuds, as you walk your dog in Amsterdam, Berlin or Prague, that sentence goes flat. You cannot see the tables. You cannot pause to look up “moderated.” You just hear difficult words about a thing you experience every single day at work, at home, in the news.

And there is a second, more delicate issue. Academic careers are built on publications, and publications are written for other academics. That is not a character flaw; it is the incentive structure of the profession. But it does mean that when I ask a scholar a question, the answer sometimes arrives pre-formatted as a journal abstract.

What the practitioners gave me

Slowly, without planning it, my guest list shifted. I found myself talking less with people who study trust and more with people who stake something on it: CEOs, entrepreneurs, diplomats, journalists, philanthropists, negotiators, consultants, psychologists.

And something happened to the conversations. They became audible.

A CEO describing the moment she had to trust a subordinate’s judgment in a crisis does something no seven-point scale can do: she makes you feel the weight of it. A founder telling you about the handshake deal that went wrong teaches you more about vulnerability, the beating heart of every academic definition of trust, than the definition itself. These stories are not data. They would never survive peer review: N = 1. Yet they travel straight from one human being to another, in any accent, at walking speed.

That, I have come to believe, is what a podcast is for.

And there is a public dimension to this, too. The word trust is now used everywhere and often carelessly - in politics, in advertising, in the daily news. In the Netherlands, where I live, politicians use it several times a day, mostly as a loose synonym for belief. A validated survey scale, however precise, does not compete for attention in that noise. A well-told story does. When a guest describes the exact moment she chose to rely on another person without a guarantee, listeners recognize its shape in their own lives, and the word begins to recover its meaning for the people who use it every day.

Not a farewell, a rebalancing

So am I done with the scientists? Absolutely not. Some of my favorite episodes were with academics, and you can hear why: they were the translators. Roger Mayer explains trustworthiness through the mechanic he chooses to work on his car.[2] Sim Sitkin, working with the design firm IDEO, shows that well-designed bureaucracy can build trust rather than corrode it.[3] Reinhard Bachmann on why so many mergers and acquisitions quietly fail because trust between the partners was never built in the first place.[4] Guido Möllering on the leap of faith that sits at the heart of every act of trusting.[5] Antoinette Weibel, herself a former FINT president and one of the field’s sharpest internal critics, is openly calling for what she terms engaged scholarship.[6] Joe Hamm is showing, through the Dow Chemical case in Midland, Michigan, how deep trust in an institution can quietly lead people to underestimate real risks to their own health.[7] These are researchers who work with real organizations, who are at least as interested in your question as in their last paper.

What I am done with, gently, is the untranslated version. Going forward, when I invite researchers, I want to ask them different questions. Not “what did you find?” but “what did the questionnaires miss?” Not “explain your model” but “when did your own trust get broken, and did your theory help you?” I suspect the honest answers are more interesting than the published ones, and I suspect quite a few scholars are quietly longing to give them.

And to my friends at FINT and beyond: consider this an open invitation, with one condition attached. Leave the regression tables at home. Bring a story instead.

Trust, after all, does not live in datasets. It lives in the moment someone decides to rely on someone else without a guarantee. My job, as I now understand it after six years, is simply to record what that moment sounds like.

---

TrustTalk is a podcast by Trust Media Foundation, a non-profit organization based in The Hague, the Netherlands, that explores trust in business, politics, and society. You can find all episodes at trusttalk.fm.

---

[1] Discussed in the TrustTalk interview with Benjamin Ho (2023). Ho, himself an economist and game theorist, notes that formal training in economics tends to push people toward the uncooperative option in trust games — not because economists are personally selfish, but because the discipline teaches rational self-interest as a baseline assumption. See also his book Why Trust Matters: An Economist’s Guide to the Ties That Bind Us.

[2] TrustTalk interview with Roger Mayer (2024). Mayer illustrated the ability–benevolence–integrity dimensions of trustworthiness through the everyday example of choosing a mechanic to work on his car — how competent do I judge them, how much do they care about me as a customer, and do they cut corners or want to do the job right.

[3] TrustTalk interview with Sim Sitkin (2023). Sitkin described IDEO’s practice of “self-appointed adults” stepping in during creative work — a formal control mechanism that strengthens rather than undermines trust because the adults are themselves creative types who share the underlying culture and values.

[4] TrustTalk interview with Reinhard Bachmann (2022). Bachmann pointed to the high failure rate of acquisitions as one of the clearest empirical illustrations of missing trust between organizations, compounded by the difficulty of integrating corporate cultures across the merging parties.

[5] TrustTalk interview with Guido Möllering (2021). Möllering drew the “leap of faith” concept from Georg Simmel’s century-old observation that trust means believing in another without being able to articulate exactly what one is believing.

[6] TrustTalk interview with Antoinette Weibel (2024). Weibel introduced the term engaged scholarship as her preferred model for the researcher–practitioner relationship: defining research questions together from the outset rather than researchers presenting finished findings to practitioners after the fact.

[7] TrustTalk interview with Joe Hamm (2024). Hamm’s study of Midland, Michigan found that residents who most trusted Dow Chemical were more likely to consume local bottom-feeding fish carrying dioxin contamination — their trust in the institution translated directly into higher-risk personal behavior.

---

[1] In social science, construct validity refers to how well a questionnaire actually measures the abstract concept it claims to measure - trust, in this case, rather than adjacent things like optimism or agreeableness. Establishing it involves showing that scores on the instrument behave as the underlying theory predicts, and don’t overlap suspiciously with measures of neighboring concepts. Essential methodological work; also, as a phrase, spectacularly hard to love.