2.1 Trust as an Individual Attitude

The dominant sociological approach to trust treats it as a subjective attitude of an individual trustor, often resting on a minimally rational assessment of the trustworthiness of a trustee (Misztal, 1996; Schilke et al., 2021). In what follows, we point out some of the problems with this approach, especially as it is encapsulated in the measurement device it employs, namely the 5-point Likert scale. Yet, we list these issues not simply as methodological problems affecting the measurement of trust, but as potential evidence about the nature of trusting as practical action. As such, they offer a rich source of insights to inform the reconstruction of the sociological theory of trust.

By asking individuals to mark “how much” they trust some entity, the Likert scale constructs trust as a more or less stable attitude held by the individual about the reliability of the trustee. If one wants to know if trust exists, one merely needs to ask the respondent to “look inside” and report on their subjective sense. This also means that trust is distinct from action and precedes it as an anticipation of the risk involved. The Likert scale, finally, constructs trust as a monotonic function that varies inversely with mistrust. When the value decreases, trust weakens and mistrust strengthens until one replaces the other.

The most conspicuous problem bedeviling such studies is the extent to which responses are highly sensitive to the wording of survey questions (Eyal, 2019, pp. 46–53). We already saw how respondents reacted positively to “my doctor”, while their actual opinions about doctors were far more negative. This is often discussed under the heading of “social desirability bias”. Democrats in our survey marked “a lot of trust” in the CDC, no doubt because by now mistrust in the CDC is politically coded as “conservative”.

This sensitivity to the cues embedded in wording means also that when a survey asks about trust in “X”, it is not given that “X” is a discrete entity in the minds of respondents. They may, in fact, be answering about their trust in “Y”. British respondents to a 1998 Monsanto survey expressed high levels of trust in GMO foods, but when a second wave of the survey prompted that “the British government has stated that it was satisfied that the product was safe,” their trust levels tanked (Millstone & van Zwanenberg, 2000). In the second wave, the public was answering about their trust in Y (the government), rather than X (GMOs).

Ultimately, this sensitivity to the wording of questions may mean that what people say in their answers to trust surveys has very little to do with what they actually do. People may mark “do not trust at all” in a survey question about “government scientists”. Yet every morning, faithfully, like their morning prayer, they count their pills and take their medications approved by government scientists at the FDA. Which should count as their true level of trust? Answers to surveys vary over time. Behavior is slower to change, but is it trust?

Respondents’ sensitivity to how survey questions are worded testifies that trusting is highly dependent on context (Lewis & Weigart, 1985, p. 976). More precisely, it shows that respondents treat the wording of survey questions as clues to figuring out the relevant context for trusting, inclusive of who should be treated as the trustee. This is why the trustee designated by survey questions may not be a discrete entity in the minds of respondents. The wording of survey questions acts as a frame that focuses attention on a particular entity. “My doctor” works in this way. Individuals sketch into these two words their image of a reliable, benevolent, familiar figure who could serve as a stand-in for what is outside the frame, namely the network of medical expertise. On the other hand, telling the British public, in whose memory the “Mad Cow” scandal is still fresh, that “The British Government” has declared GMOs safe, shifts the frame and brings into view a morally polluted entity that thereby infects GMOs. What the “problem” of wording reveals is the interplay between frames, the networks that lie outside them, and individuals’ search to discover, in the details of situations, in whom to lodge their trust.

The same analysis explains the discrepancy between trusting-as-action and trusting-as-answering-a-survey. The frame within which taking a pill takes place does not focus attention on the ultimate guarantor that lies at the other end of the network. It is a frame of “familiarity” (Luhmann, 2017, pp. 21–26), focusing attention on details that are on this end, like what the label on the bottle looks like, how was it obtained, what we think about the doctor who prescribed it, etc. The survey question changes the frame and with it the contextual clues for who is the relevant trustee.

Talking about “social desirability bias” begs the question: what self are respondents seeking to project by indicating their trust or mistrust? Clearly, they try to present themselves as responsible, neither trusting blindly, nor paranoically refusing all trust. What the “problem” of social desirability bias reveals is that communications of trust and mistrust are part of the “forensic vocabulary” (Douglas, 1990) by means of which we hold each other accountable and seek to persuade ourselves and others that we are trusting responsibly.

Finally, the construction of trust in this approach as a monotonic function is contradicted by what we all know, namely that if we trust someone a lot and are disappointed, our trust will flip into its opposite, a sense of betrayal, in non-monotonic fashion. One answers “do not trust at all” (1) today, precisely because one answered “trust a lot” (5) yesterday. This is extremely instructive. It shows that trusting is asymmetrical and eventful: “[Trust] is typically created rather slowly, but it can be destroyed in an instant — by a single mishap or mistake” (Slovic, 1999, p. 697).

These issues reappear also in more conceptually sophisticated sociological studies, such as Meyer et al.’s (1995) model of organizational trust, which has been cited 32,677 times and enjoys broad cross disciplinary consensus (Schilke et al., 2021, pp. 240–241). This is because treating trust as a subjective attitude measurable by a survey instrument means that they try to resolve by definitional fiat problems that actors negotiate in practice. As we saw earlier, the survey question requires identifying in advance an entity that serves as the trustee. Accordingly, Meyer et al. (1995, p. 711) restrict trust to a relationship between two clearly identifiable parties, trustor and trustee (see also Robbins, 2016, p. 978; Schilke et al., 2021, p. 241). A farmer planting a crop, they say, is not trusting because there is no relationship with an identifiable trustee. The weather forecasting system, they say, is not a trustee, because “meteorologists do not control the weather” (Meyer et al., 1995, p. 725). What was billed as conceptual rigor becomes a dispute with the actors about what entity can legitimately be assigned responsibility as trustee. The farmer may consider the weather forecasting system as a trustee (indeed, crop insurance contracts may require this), or she may designate God as the trustee. After all, God does control the weather. In short, Mayer et al.’s insistence that trust can only exist when it has a clear dyadic form prevents them from studying how individuals search — in real time, attentive to situational details, without a scholastic warrant on what is “really real” — how to give their trust a dyadic form.

Similarly, the survey situation requires that they will define trust as an explicit recognition of risk, “the willingness of a party to be vulnerable to the actions of another party,” thereby excluding from consideration trusting that is tacit and routinized (Meyer et al., 1995, pp. 712–713; Schilke et al., 2021, pp. 244–245). Schoorman et al. (2007) operationalize this as agreement with the proposition “I would be willing to let my supervisor have complete control over my future in the company” (p. 352). It is quite possible, however, that prior to being asked this question the employee has never consciously entertained assuming this “risk”. It is also possible that things at the company are arranged so that the supervisor already has a lot of control over the employee’s future. The employee’s attitude seems more like “tacit acceptance of circumstances in which other alternatives are largely foreclosed” (Giddens, 1990, p. 90), which in Meyer et al.’s approach is not trust. Yet, it becomes trust solely by virtue of being asked by the researchers.

2.2 Trust as a “Leap of Faith”

Outside the enterprise of trust measurement, most sociological theorists of trust do not think of trust as a subjective attitude. A longstanding tradition in sociology, reaching back to Simmel (1978), and further developed by Luhmann (2017), Lewis & Weigert (1985), Giddens (1990), Möllering (2006) and Sztompka (2006), seeks to uncover the deeper structures underneath the contradictory ways people talk about trust. If what people do contradicts what they say, perhaps we should ignore what they say and treat trust as “a collective attribute […] applicable to the relations among people rather than to their psychological states taken individually” (Lewis & Weigart, 1985, p. 968).

In direct contradistinction to surveys, what is common to this group of theories is that they tend to marginalize what individuals say about why they trust, and distinguish it sharply from “trust itself”. Simmel (1978) said that trust “may rest upon particular reasons, but is not explained by them” (p. 178). Luhmann (2017) agreed: people are “never at a loss for reasons […] of why [they] show trust in this or that case” (p. 29). Their accounts, however, cannot really be why they trust, because “no decisive grounds can be offered for trusting.” The trusting individual confronts the need to choose a course of action without ever being able to weigh the full complexity entailed by alternative contingent futures (Lewis & Weigart, 1985, p. 970). Möllering (2006) agrees and says that the actors’ post-hoc “rationalization […] obscures the leaps of faith which the actors […] still make if they genuinely trust” (p. 118). In interviews, “it may be necessary to discourage respondents from over-rationalization and to probe specifically for references to information that trustors consciously miss or dismiss” (p. 119).² Giddens (1990) simply says that “all trust is in a certain sense blind trust!” (p. 33).

The reasons an individual gives are meant “to uphold his [sic] self-respect and justify him socially” (Luhmann, 2017, p. 29). Essentially, Luhmann is saying that trusting involves skillfully convincing oneself and others that one is trusting responsibly, not blindly, but that the reasons provided “are brought into account for the placing of trust, but not for trust itself” (ibidem).

This is a peculiar formulation. What is “trust itself” as distinct from the “placing of trust”? This distinction is made by all the aforementioned theorists (Möllering, 2006, p. 119; Solomon, 2000, p. 234; Giddens, 1990, p. 99; Lewis & Weigart, 1985, p. 973) and is what allows them to discount what individuals say about why they trust. Their writings offer three interrelated rationales for this.

First, trust is inescapable and obligatory like a Durkheimian social fact. Suspended in midair at 40,000 feet altitude, we trust — in the pilot, in the aircraft, in the Boeing company, in the FAA. Do we really have a choice? Trust in this situation is “tacit acceptance” (Giddens, 1990), which reduces complexity by essentially proceeding “as if” this complexity did not exist (Luhmann, 2017, pp. 27–35; Möllering, 2006, pp. 111–115). The function performed by trust testifies to its collective nature. It is the “mutual faithfulness” on which all social relations depend (Simmel, 1978). Each individual trusts by observing that others trust, or as Lewis & Weigert (1985, p. 970) put it, the cognitive basis of trust is “trust in trust” (see Luhmann 2017, pp. 73–78). Nobody trusts alone and nobody can go about their lives without trusting. Even mistrust simply means trusting someone else. Hence the reasons people cite for their trust can be ignored (Luhmann, 2017, pp. 53–67; Lewis & Weigart, 1985, pp. 968–969; Giddens, 1990, pp. 21–30).

This line of reasoning is insightful and a strong antidote to attitudinal theories of trust. At the same time, it means that the most important questions about trusting, the questions of change or variation, are not addressed: Why is trust placed in x over y? Why is trust replaced by mistrust, and when? Why do some people mistrust what other people trust? These questions are not answered by asserting that trust must exist because it is functional.

Trust differs from other mechanisms for reducing complexity because the “missing information” is replaced by “internally guaranteed security” (Luhmann, 2017, p. 103). This is the second rationale functionalist theory provides for why it discounts what individuals say. All the aforementioned theorists locate “trust itself” at an internal, emotional level, where it exists independently of the reasons given for it. Simmel (1978) said that in trust there was “an additional element that is hard to describe […] an assurance and lack of resistance in the surrender of the Ego to this conception” (p. 178). Luhmann (2017, p. 31) explained this aspect of trust as something that is learned during early childhood, when infants and toddlers learn “how to trust”, and develop a sense of themselves as people who would honor the trust of others. Möllering (2006) also speaks of trusting as a “skill learned in infancy” (p. 117). Giddens (1990) draws on object relations theory to describe “basic trust” as an “extremely sophisticated methodology of practical consciousness” (pp. 92–100), which develops in infancy. Finally, Lewis & Weigart (1985) echo Durkheim’s theory of solidarity when they describe trust as arising from an “emotional bond among all those who participate in the relationship,” and leading to “intense emotional investments” (p. 971). Trust is sustained, therefore, by anticipation of the “severe emotional pain” that its violation will inflict on all involved.

We owe much to functionalist theory. We have learned from it that trusting is a skill acquired early in life, and consisting of a “methodology of practical consciousness”. Moreover, Lewis & Weigart’s (1985, p. 971) argument about emotional investment in the object of trust can explain the asymmetry and non-monotony of trust — the aspect that bedevils attitudinal theories — through an analogy with Durkheim’s theory of the sacred. Like the sacred object, the trustee is an emotionally invested symbol (Luhmann, 2017, p. 32) built through collective rituals that hark back to the development of basic trust in early childhood. Yet, like the sacred object that can be contaminated by the fleetest contact with the unwashed, “one falsehood can entirely upset trust […] [and] unmask the ‘true character’ of somebody” (Luhmann, 2017, p. 32). The trustee becomes a polluted object because the breach of trust leads to a “flooding-in of existential anxiety” (Giddens, 1990, p. 98).

Nonetheless, locating “trust itself” at the emotional level is not tenable. “Taken to extremes,” say Lewis & Weigart (1985), “if all cognitive content were removed from emotional trust, we would be left with blind faith” (p. 972). Their solution is to say that “trust in everyday life is a mix of feeling and rational thinking.” This compromise is in fact a Trojan horse. When they add “good rational reasons” to “positive affect”, they essentially sneak back in the point of view of the actors and their justifications. If people are “never at a loss for reasons”, would the sociologist now classify some of these reasons as false consciousness and others as truly rational, in order to distinguish “trust itself” from blind faith? If not, it means that the sociologist always has to consider what people say when they try to accomplish trusting in practice.

The third rationale employed by functionalist theory to discount what people say about why they trust has to do with time. Settling into our airplane seats, we could come up with all sorts of reasons why it is rational to trust the airline, but each of these reasons can be countered with further skepticism — have we actually seen the technicians inspecting the aircraft? This game of reasons and counter-reasons is an infinite regress, and nobody has enough time to go down the rabbit hole of alternative contingent futures (Luhmann, 2017, p. 28). It follows that the only way to “reach the state of trust as such” is at a certain point to take a “leap of faith”, which is “the true essence of trust” (Möllering, 2006, pp. 105–106; Guthrie, 2008, p. 134).

This argument has been most consistently developed by Möllering (2006, pp. 105–126), though he shows that all the other theorists also refer to some idea of a “leap of faith” (p. 110). Trusting is achieved by “suspension”, for which the leap of faith is “a very apt image” (ibidem). Suspension is like phenomenological “bracketing”, a sort of “nevertheless” through which “actors manage to live with the fact that there are gaps and missing pieces” (p. 115). It is “an extremely sophisticated methodology of practical consciousness […] a tacit, continuous monitoring of the normality of the situation” (p. 117).

Once again, we are indebted to Möllering. The great advantage of his approach is that he focuses on the process by which “the state of trust is reached,” and shows how reasons, routines, reflexivity and interaction all play a part in this process. His approach has inspired fruitful empirical investigations (Brownlie & Howson, 2005; Guthrie, 2008).

That said, there is a glaring contradiction between characterizing suspension as “continuous monitoring of the normality of the situation” and the image of a “leap of faith”. The leap of faith is a peculiar temporal metaphor. It describes a moment when time slows down to a standstill, even reverses (“his whole life flashed before his eyes”), only to become condensed, like a black hole, into a single moment of instantaneous decision. It is very far from the ordinary experience of trusting, and there is nothing about it that involves “continuous monitoring”. Möllering (2006), seems to recognize this problem because he qualifies the image and speaks of situations when individuals “can only take one very small leap of faith after the other” (p. 118), which sounds more like continuous monitoring. Yet, in the same breath he takes Kierkegaard’s interpretation of the biblical story of the Akedah as “reveal[ing] very strongly the essential meaning […] of the leap of faith” (p. 117). Instructed by God to sacrifice his son, Abraham eschews any form of justification and chooses blind faith in the absurd.

Möllering defends this discrepancy by saying that his point is simply “to clarify when it is justified to speak of trust and when it is not” (p. 118). Namely, the contradiction results from determining by definitional fiat which and when is “trust itself”, instead of attending to the diversity of “methods of practical consciousness” that actors draw upon to determine this in practice. Most importantly, the contradiction results from a “scholastic fallacy” (Bourdieu, 2000). The leisure of the theorist is what permits her to stop time in its tracks so as to follow all the reverse paths on the decision tree and counter all the “good reasons” with “other good reasons” (Guthrie, 2008, p. 134). This is what justifies concluding that the reasons provided by the actors are not related to “trust itself”. The characterization of trust as a leap of faith reflects a point of view outside the action; a point of view which has all the time in the world to show, by contrast, that the actors can never have enough time to trust reasonably, but must be taking a leap of faith. In ordinary language, however, we distinguish responsible trust from “blind faith”. It is a consequential distinction that people work hard to discover in the details of practical situations, attending precisely to how the passage of time is handled: was trust demanded too quickly? Has enough time elapsed for the placing of trust to be accountably responsible?

The mark of the scholastic fallacy is that the theory treats the point of view of the actors “as an obstacle to be destroyed and not as an object to be understood”(Bourdieu, 2000, p. 189).³ Put differently, however accurate the theoretical model, the actors could never take it as the basis for their action. If the recipient of a gift acted on the “truth that emerges from the model” and immediately offered a counter-gift, she would have insulted the other party and destroyed the whole delicate dance of gift exchange. Similarly, however insightful the recognition that trust ultimately involves a “leap of faith”, if the actors take it as a basis for their action, they would appear to all involved, including themselves, as placing blind faith rather than trusting.

The problem for the sociologist is not to decide by definitional fiat which of these truths — the truth of the actors’ point-of-view or the truth revealed by the model — is the real one, but how to “hold [both] together, so as to integrate them” (Bourdieu, 2000, p. 189). We can begin doing this by directing sociological attention to the mechanisms by which the actors themselves are able to hold on to both truths at once. In the case of the gift, “it is the lapse of time between the gift and the counter-gift” (Bourdieu, 2000, pp. 191–192), the actor’s practical sense of timing, and the institutionalized temporal frames that organize what time horizon is foregrounded.

Research on trusting requires a similar approach that focuses on the mechanisms actors employ to manage the contradiction between the subjective experience of trust as well-founded and the scholastic demonstration that trust is an absurd leap of faith. Actors are in fact aware of the tension, and refer to it obliquely through forms of gallows’ humor. What we mean by “trust methods” is an extension of Möllering’s characterization of suspension as “an extremely sophisticated methodology of practical consciousness,” but we think there are multiple such methodologies, namely practical devices that people employ to accomplish trusting as accountably responsible (Garfinkel, 1967, p. 1; Heritage, 1984, pp. 142–159). If we abandon the search for “trust itself”, the question becomes not whether one has taken a leap of faith at one point, but how trusting is accomplished and maintained throughout? How and why it grows, diminishes or breaks down? To answer these questions, we do not need to accept people’s stated reasons at face value, yet they are valuable as demonstration of how trusting involves a search for the proper context within which trust would be accountably responsible.

To summarize, once we set aside the theory’s attempt to pin down “trust itself”, we can convert its insights into a more flexible analytical framework for the study of trusting. First, suspension no longer appears as the “essence of trust” (Möllering, 2006, pp. 115–119), but as one trust method among many. People may say that they “had to take a leap of faith,” but they typically mean by this what Möllering characterized as taking “one very small leap of faith after the other,” namely displaying to oneself and others that one is aware of having taken such a leap and is hedging and monitoring to make sure it was justified.

Second, the emphasis on the emotional dimension of trusting reveals that the problem addressed by trust methods is not whether to trust or not, but how to trust in a way that is accountably distinct from blind faith and debilitating mistrust. Trust methods are an integral part of the everyday rituals, at once emotional and practical, by which the sacred self is constituted and protected (Goffman, 1971). That’s why trusting is asymmetric and non-probabilistic, because it is about the moral status of selves.

Third, the focus on trust as a collective social fact reveals that the skilled act of trusting is typically supported by institutionalized frames that foreground a particular figure of the trustee, and relegate the infinite regress of reasons and counter-reasons to the background. Trust methods are typically adjusted to these frames, and hence appear to require no skill. Yet, in less structured situations it is possible to observe that they consist of a search, in the details of the situation, for how to trust responsibly.

Finally, the claim that actors do not have enough time to trust reasonably reveals — once we discard the scholastic fallacy — that attention to temporal variables such as duration, interval, timing, sequence and speed is at the core of trust methods.

3.1 The Study

The quotes below are taken from a study of long COVID patients conducted in 2021 and 2022. Individuals responded to posts on Long COVID patient groups on social media platforms. They took a survey that included an open-ended set of questions and were asked whether we could contact them again for an interview. The interviews were semi-structured and designed to elicit how respondents speak about who, what, why, and how they trust. The quotes below are from 382 survey respondents residing in the United States — with particular attention paid to the 334 with Long COVID — and 91 interviewees with Long COVID.

The sample we obtained was not representative of the US population, nor likely the Long COVID population. Our 382 survey were generally more likely to be female (71%), white (76%), college educated (68%), suburban (50%), employed (55%), and less likely to be conservative/Republican (14%). These skews were the result of class differences in access to social media, preponderance of women and absence of racial diversity in the forums where we advertised, and political polarization that predisposed conservatives to mistrust our survey.⁴ These skews need to be kept in mind when interpreting the findings.

3.2 Distinguishing Responsible Trust from Blind Faith

Trusting is inescapable. Hence the skill involved is concerned not with whether to trust or not, but with how to trust in a way that is accountably distinct from blind faith and debilitating mistrust. This was particularly evident in how our interviewees spoke about evaluating the trustworthiness of information they found online. Recall that in Table 1, “my own research” was ranked most trusted by survey respondents. When we asked interviewees about why and how they trusted their own research, it became clear that they were keenly aware of the moral panic about people who “believe everything they read on social media.” They were concerned to distance themselves from this image. When asked if they worry about online misinformation, interviewees said that they “worry about the general public” (Interview #88), but not about themselves because they employed multiple checks and cross-references. They said things like: “I get at least three or four other sources that are not related to each other. […] really good sources and peer-reviewed articles.” (Interview #21); or “If it’s from a journal, I check out the journal and who owns it. And you know again the studies, like the sample size. […] If it’s from one of my support groups where is just patients reporting what’s working and what hasn’t worked, I’ll go look up further information from studies” (Interview #88).

In short, as interviewees described how they conducted online research, they were concerned to present themselves as different from “the general public”, who, in their depiction, believed everything they saw online. They drew a contrast between blind faith in online information or in what “just patients” said, and their own responsible trust, bolstered by an elaborate set of checks.

One possible objection is that this merely reflects the preponderance of the highly educated in our sample. This objection cannot be dismissed until the research is replicated in a different sample, but our hypothesis is that while details may differ, we are likely to see a similar concern to distinguish blind faith from responsible trust. Less educated respondents might draw on a different set of resources in doing so — they might consult a friend rather than JAMA; they might cross-reference with a source that we think is unreliable — but they would be similarly concerned to draw a distinction, and our approach allows to treat their trust methods symmetrically.

Indeed, many of our respondents said that “the only ‘trustworthy’ information about Long COVID is the symptoms that people describe” (Survey Respondent #215). This can sound like blind faith in whatever other patients say because “they have firsthand experience vs. doctors that have no knowledge” (Survey Respondent #47). Our respondents, however, accounted for this trust as responsible by attending to temporal variables. First, trust in other patients’ reports was explained as reflecting the newness of Long COVID, and the lack of reliable medical information about it because scientific research “takes a long time” (Survey Respondent #244), and doctors were “about 50 steps behind” patients (Survey Respondent #277). Differences in speed account for why this respondent, who “normally trusts doctors”, is “embarrassed to say I put so much stock in a Facebook group, but honestly there was nothing else out there” (Survey Respondent #277). Second, while expressing trust in other patients’ reports, they qualified it by taking an initial stance of charitable interpretation: “When people say they’re having unusual symptoms, I tend to give them the benefit of the doubt because there is a lot we still do not know” (Survey Respondent #205). Giving the benefit of the doubt is different from blind faith. It essentially says that “what works for one person may not work for another” (Survey respondent #109). Trusting in other patients’ reports is therefore qualified as “until further notice”, a matter of sequence. It is a responsible stance to take initially, but it can be withdrawn later. Indeed, our respondents also described how after reading “about other people’s symptoms compared to yours”, the next step would be to “still do personal research through medical journals and talk to my doctor” (Survey Respondent #47). Or when a particular medicine is discussed “on one of the support groups”, they would follow up by “going to […] the source, not you know, person X in Denver, Colorado with Long COVID, but […] which scientists and doctors is this emanating from? […] the medical articles that they have published in credible medical journals” (Interview #2).

Another possible objection would be that this concern to distinguish their trust from blind faith reflects social desirability bias compounded by the known pitfalls of retrospective reconstruction. Respondents are telling a self-legitimating story, but does it accurately represent why they trusted? They probably took a leap of faith, and now in retrospect they are trying to present it as rational. Empirically addressing this objection requires a longitudinal research design. We will discuss this in the conclusions. At the same time, it should be clear from the preceding that we think of trusting not as a one-and-done, but as a process in which there can be slowdowns, reversals and forks. Individuals’ reflexive accounting of this process plays a crucial role in organizing it.

3.3 Trust Methods Supported by Institutionalized Frames

Left to their own devices, it would be exceedingly difficult for individuals to draw the line between blind faith and responsible trust. As is clear from the previous quotes, our interviewees displayed a distinct preference for orthodox markers of scientific credibility. Asked what makes information found on the web trustworthy, they answered: “published, scientific medical journals containing studies done by credited professionals” (Survey Response #100) or “verifiable scientific research based from a university or hospital study” (Survey Response #97). Since these markers are often listed in theoretical analyses of “trustworthiness”, it is important to emphasize that this preference is by no means universal. As noted earlier, conservatives shunned our survey precisely because it was affiliated with a reputable Ivy League university. Preference for these markers is a form of social positioning and reflects our interviewees’ pathos of distinction: “Because I have my two master’s degrees, I know how to research scientific and peer-reviewed articles” (Interview #101; Nurse Practitioner); “Being able to read studies and read stuff […] kind of allowed me to […] be able to look at information in a healthy way” (Interview #5; Math Major). 

These orthodox markers of scientific credibility do not convey trustworthiness by themselves, but only when viewed from within the relevant frame. Our interviewees conducted Google searches and carefully read the results, but they drew on their education and/or professional training to frame this activity as “research”. This frame contains an institutionalized hierarchy of credibility, and thus allowed interviewees to account for their trust as a process of climbing this hierarchy from least to most trustworthy: “If I see something online often it’s on a Facebook page or something. I’ll look at that and I’ll run it through a Google search, and until I see it from a JAMA or ATS or Chest or Lancet […] I don’t usually put a ton of faith in it” (Interview #31). Even within this highest rank of credibility, some respondents made fine distinctions. They wanted to see that researchers have “a documented history of reputable work.” That they are known to “reference or listen to Long Haulers”, and are not the sort of researchers who “run headlines that so and so has found a cure or drug that solves Long COVID and then states that findings are based off mouse model experiments and need further validation” (Survey Response #165). Note how this respondent’s familiarity with medical research allows them to frame their mistrust as an integral part of the process of trusting.

3.4 Trust Methods Are Not Pregiven Scripts but Situated Practices

While the preceding is a useful correction to media depictions of ordinary people being fooled by peddlers of misinformation, it runs the risk of making trust methods appear like pregiven scripts that individuals can simply follow to assess the credibility of information. This was not our intention. We cited the elaborate checks employed by interviewees not as evidence for what are trust methods, but to show that trusting involves drawing a distinction between responsible trust and blind faith. Now we would like to show that how to trust responsibly is not given in advance, but must be discovered in the details of the situation. Trust methods are not scripts in the way that our interviewees self-confidently depicted them, but “situated practices of looking-and-telling” (Garfinkel, 1967, p. 1). Moreover, discovering how to trust responsibly is not a one-off. Situations ramify and create new perplexities, which means that discovering how to trust responsibly is a continuous task, “an endless, ongoing, contingent accomplishment” (Garfinkel, 1967, p. 1).

Recall how interviewee #31, who is an ICU nurse, said that she would trust online information only if she could trace it to a prestigious outlet like JAMA. What would happen, however, if another article in JAMA reported contradictory findings? Which one should she trust? As another interviewee commented, “it’s hard to know when they contradict each other, especially when some of the sources that contradict each other are pretty good” (Interview #33). She could read further and follow the citation trail, but she may quickly discover even more unresolved disputes.

One of our interviewees summed up this experience aptly: “And so they send you down, you know, the research rabbit holes, you know, reading as much as you can. But that was a really, really big one for me because it was something I could have hung on to” (Interview #21). By “research rabbit hole” she is referring to the theory that Long COVID is caused by activation of a dormant Epstein-Barr virus (EBV). Because she had mono as a child, she thought this theory could be true and apply to her. However, the theory remains highly contested, and some studies have found no link between the two illnesses. The interviewee was not unprepared to delve into the debate. She had an undergraduate degree and her husband was Chief of ER medicine at the local hospital. However, there was reason for concern. Wouldn’t she appear as one of those people who trust whatever new-fangled theory they read on the Internet? Then she joined a Facebook group for Long COVID patients. In the group, she interacted with several women, who, like her, were middle-aged, had contracted mono in adolescence, and now suffered from Long COVID. This is what she means by “something I could have hung on to.” It is equivalent to asking rhetorically, “What are the chances that there would be this kind of coincidence?”. Once she felt reassured that to trust this theory would not be blind faith, she asked her doctor to test her for EBV. The test returned positive and showed “off the charts” signs of EBV.

This story is instructive for the following points:

The expression “research rabbit holes” bears striking similarity to what functionalist theorists identified as the infinite regression of reasons and counter-reasons that precludes explaining trusting by reference to any “decisive grounds”. This means that the scripts that our respondents described must be seen as only the thin outer layer of their trust methods. Trust methods are sedimented. They are organized in several superposed layers that an experienced interviewer must probe, because the real problem for trust methods is how to stop the descent into infinite regression, whether it is a “research rabbit hole” or, as we shall see below, deciding whether other patients are “like me” and can be trusted. As a result, in any given situation, there would be several devices employed and linked together.

Interviewee #21 says that she needed “something I could have hung on to.” The imagery is not of a “leap of faith”, but of an anchor, something solid enough to stop the spiral. Her use of the conditional “could have” indicates that she is reflexively aware that it is not a “decisive ground”, yet it is a “big one”, namely it is not arbitrary, it makes sense of her present circumstances and allows her to proceed. It is defensible by asking rhetorically “what are the chances?”. In short, the problem is not just how to stop the descent down the rabbit hole, but how to do so and yet trust responsibly. Individuals need to discover, in the details of the situation, the means for doing so.

The “big one” for interviewee #21 was the encounter with “patients like me”, who shared her symptoms, her social position, and her illness history. This can be interpreted as a simple “tribal” script — trust only people who are the same as you. This interpretation, however, ignores that who is a patient “like me” is not obvious. When reading other patients’ descriptions of their symptoms and what worked for them, it is not obvious how to determine which ones are “the same” because patients’ reports contradict one another: “There’s a lot of noise out there in the Long COVID space, and it’s hard to know who to trust” (Interview #5). “You just like read all these like posts, but you’re not sure how reliable they are” (Interview #1). This is the same problem as with “research rabbit holes”. It leads to a similar search for situational details that can confirm that one is trusting responsibly. Trusting other patients is a situated practice, not a ready-made script.

Many interviewees tell a story of how they came to trust other patients, which is equivalent to interviewee #21 finding that other patients had mono too. They compare symptoms, but they give special status to symptoms that are unusual. When “you say, ‘it feels like I’m walking on a boat.’ And 50 other people say, ‘Yeah, that’s exactly what it feels like.’ Well, then you know, you’re not alone. You didn’t make it up in your head” (Interview #5). One of our interviewees experienced hallucinations. She found it “extremely reassuring” that the nurse could tell her from her own experience that this happened to others as well, “when, you know, we know doctors could tell us […]” (Interview #11). She didn’t complete the sentence, but the meaning was clear: doctors would tell her that she did “make it up in her head.”

The fact that “some of these symptoms are so wacko at times” (Interview #21) offers a quasi-rational support for recognizing one another as “like me”: two very low probabilities multiplied make for an even lower probability that the two patients were not taken from the same pool. This is rational reasoning in the sense that Garfinkel (1967, pp. 262–283) speaks of the “rational properties of commonsense activities”, but it cannot be applied generally to all situations. One has to find in the situation the means for its applicability, among which is placing it in the proper temporal sequence. “Wacko” symptoms work to link patients to one another not only because of their coincidence, but because they embed patients in the same narrative sequence. These symptoms were a chief reason why patients were disbelieved by doctors and felt that they were “gaslit” by them (Au et al., 2023): “We all basically found each other because we’ve all been fighting the system” (Interview #21). The narrative arc begins with betrayal and self-doubt, and proceeds to finding validation and collective identity: “It’s an opportunity to just kind of share similar symptoms and helps us to not feel so alone” (Interview #21).

Trusting by sharing strange symptoms is not a script, but situated practice. Symptoms and stories may differ in subtle ways. Accepting them as “the same” and trustworthy depends on attention to situational details and sequence. What if the person who had symptoms similar to you also reported that their symptoms were relieved by taking Ivermectin? Many of our interviewees told us that they withdrew from participating in online Long COVID groups because they did not trust many of the participants, especially about “what you should or shouldn’t do.” At this point, they begun to wonder whether they really shared symptoms with these other patients. “It’s just such a wild west,” the only responsible stance would be to “kind of ignore [all] that because, you know, you don’t really know” (Interview #81). They made situational discriminations based on their sense of what trusting other patients would signal about how trustworthy they themselves were. Some declared that they only trusted “people in the Long COVID community that are science-based” (Interview #5), thereby characterizing themselves as “science-based” and as “completely filtering out” other patients who were not (Interview #30). But even “science-based” patients would be distrusted if they “speak in absolutes, or unsympathetically” (Survey Respondent #109), because speaking in absolutes essentially means skipping over the necessary first stage of giving the benefit of the doubt. Finally, patients also scrutinized the situation for what trusting would signal about their political identity. One of our interviewees was prescribed Ivermectin by a clinic that he initially trusted. He was “really conflicted […] really torn,” but ultimately decided not to take the drug because “I subscribe to the liberal doctrine […] [and] this was the medicine that the dum-dums were taking” (Interview #99).

3.5 Who Is the Trustee Is Not Given in Advance, but Determined by Trust Methods

Unlike surveys that ask respondents “how much do you trust in… X; in… Y; and in… Z,” our interviewees constructed the trustee, the object of trust, by moving fairly freely from X to Y to Z and building linkages among them. Trust in the research about EBV was accountably responsible by adding to it the encounter with patients “like me”. Interacting with other patients was especially helpful when they were “reporting out what their doctor or their Long COVID clinic saw or told them.” It was even more “reassuring” when the reports came from “people all over the world, so someone’s like ‘hi I’m in Denmark and I just went to the COVID clinic and this is what they told me.’” By herself, the Dane patient inspires no confidence: “I am like, ‘is this legitimate?’” But taken together, “hearing that other Long COVID clinics are saying it too is really helpful” (Interview #41). Would we say that the trustee was the Dane patient? Or the COVID clinic in Denmark? Or the forum where one encountered this patient? Or the people who referred one to this forum? Or one’s own online research? Clearly, the trustee is none of these, but all of them together, or more precisely who or what becomes the trustee is not fixed, but a matter of changing situational determinations, as one moves along a network of expertise. Many of our interviewees said that the best thing that happened to them is that through their research they discovered online groups dedicated to Long COVID. By joining these groups, they found other patients with similar symptoms. Through these patients, they heard about a doctor they could trust. Or vice versa, a doctor they trusted connected them to another patient who connected them to a social media group. In such cases, the trust concatenated by these encounters radiated back to all involved. If something misfired, the mistrust similarly spread from one node to another.

3.6 Trust Methods Involve a Practical Sense for Timing, Sequence, Interval, Speed and Duration

Let us return to the story told by interviewee #21. A particular narrative sequence played an important role in how she accounted for her trust. The sequence began with reading about EBV, and then proceeded through falling down the research rabbit hole, encountering other patients, discovering that they share a similar biography, and finally obtaining a confirmatory test result. If we change this sequence, the quasi-probabilistic appeal to “what are the chances” disappears. For example, if the sequence began with the patients’ group, from which the interviewee was sent to read about EBV, the encounter with patients sharing her history would have lost its dramatic effect as a “discovery”. Similarly, from the point of view developed here the test result is not objective proof but an element that derives its value from coming at the right moment in a narrative sequence. If the test had not come at the end, it would have lost its status as a “clincher” providing final confirmation. Since we know she had mono as adolescent, it is not surprising that her EBV levels were high.

Attention to temporality was central also to how patients accounted for trusting doctors. In general, it was rare for them to declare unequivocally “I definitely trust doctors” (Interview #41). Typically, they did so to preface a story about a doctor that they did not trust, e.g. “a really bad cardiologist […] [who] just left me crying” (Interview #41), so they won’t be perceived as somebody who completely mistrusts medicine. The doctors that they did trust typically were ones who could confirm that they “had seen like a hundred people just like me” (Interview #41); who were “exposed to so many people with the symptoms that she could see the lay of the land and tell me like what was normative and what she’s heard was helpful” (Interview #11); and to provide “validation […] that ‘oh you’re not crazy, I’m sure these symptoms are real’” (Interview #38).

At the same time, trusted doctors were ones who also displayed epistemic humility, who were ready to “acknowledge that […] we really don’t understand a lot about this virus” (Interview #38). In contrast, doctors were mistrusted who lacked humility, who tried to “lead you to believe that they’re experts in it” (Interview #68). This would seem to set up a contradiction: doctors are trusted if they display familiarity with Long COVID, but are mistrusted if they try to pass themselves as experts. The contradiction, however, is in theory, not in practice. Trustworthiness is conveyed, and trusting is accomplished, through a practical sense for correct timing and sequence. For example, when doctors said right off the bat that they did not know enough, some interviewees did not interpret this as epistemic humility, but as unwillingness to listen (Interview #21), but if the doctor spent “quite a lot of time listening to what I was going through,” they interpreted it as epistemic humility (Interview #38). Spending “quite a lot of time”, however, could backfire. When a doctor was willing to do research and experiment with new treatments, respondents trusted them… until the sequence got too long and became like “just throw[ing] darts on the dartboard when the lights are off. You know just kinda guessing things” (Interview #51).

A display of humility won patients’ trust especially when it came in the right sequence, after a contrary experience with a doctor who displayed “hubris” (Interview #68). But if it came after a long diagnostic odyssey, it was no longer perceived as trustworthy but as one more dodge by an unsupportive medical system. At this point, a doctor could be trusted if at the right moment — at the end of a long diagnostic odyssey, and after taking time to listen — she simply said the equivalent of “trust me!”. At the right moment, “trust me” became accountable not as blind faith but as having finally found something “to hang on to”:

And after 15 minutes, she said to me, I want you to put your spiral down. You don’t need to take any more notes. You don’t need to read any more research articles. You need to let me handle it. You’re really sick and […] I believe everything that you’re saying. And you know, we’re going to get through this together. (Interview #21)

Note how interviewee #21 interprets the time that elapsed as a measure of being truly listened to. Being asked to put the spiral down is a potent symbolic acknowledgement of the burden she carried. One can almost feel the sense of physical relief this gesture conveyed. But the gesture and its phrasing in the polite imperative — “I want you” — also draws a mark in time, a point separating a “before” and “after”. This gesture could not be effective if it came earlier in the patient’s career, nor earlier in the session. It displays a practical sense for timing and a surfeit of emotional intelligence.

3.7 The Limit of Trust Methods: Play

We did not encounter respondents who professed blind faith in doctors, other patients or internet sources. No doubt there are areas of life where professing blind faith (in God, in one’s platoon buddies) is sanctioned, but dealing with Long COVID is not one of them. Nonetheless, this does not mean that respondents were always concerned to distinguish responsible trust from blind faith. Some respondents found another way in which to stop the descent into “research rabbit holes” by adopting a playful attitude. Neither distinguishing blind faith from responsible trust, nor collapsing them together, they relativized the distinction by treating it unseriously. Interviewees reported, for instance, that they tried a certain therapy or followed a particular thread of information not because they trusted it but because it did not cost much — it was unlikely to be dangerous, it was unlikely to reflect poorly on them, it did not take much time and effort:

And then some people have been like, “Oh maybe try smelling essential oils, like, maybe because they’re strong, you’ll smell it.” I was like, “Okay, okay, maybe, I don’t know. That could be fun.” I mostly have just done that to see like, “Where am I at right now?” (Interview #61)

These interviewees experimented with a variety of low-cost, low-risk remedies — from vitamins and essential oils to burning oranges — not necessarily because they trusted the advice of other patients or believed in their healing capacities. Patients often explained their playful choices by prefacing them with caveats like “take it with a pinch of salt” (Interview #100) or “I would’ve tried anything… why not?” (Interview #99). They gave themselves leeway to “try anything” because these remedies could be accounted for — to themselves and others — as harmless, cost-less, playful shots in the dark. We think about these examples as the lower limit of trust methods, so to speak. On the one hand, by treating it as “fun”, the interviewees were noncommittal about whether they trusted the advice or not; on the other hand, functionally speaking, taking a playful attitude allowed respondents to do what trust methods do: suspend their doubt and decide to go along with “anything”. Well, not quite anything:

He interviewed some respected doctors and they basically communicated a treatment plan which was mainly supplement-based so it was things like Vitamin C, Vitamin D, Quercetin, etc. That decision was relatively easy to make, like “Am I going to try this or not” because these were things that were readily available. The Ivermectin antiviral, that was slightly different because they require prescriptions. So there was different kinds of decision making. (Interview #53)

The exception that proves the rule, Ivermectin, involves “a different kind of decision-making” not simply because it requires a prescription, but because it requires taking a “serious” step like asking a doctor for the prescription and thereby abandoning the playful attitude, presenting oneself as committed, and risking the appearance of putting blind faith in unproven advice. The exception that proves the rule, Ivermectin, demonstrates that even respondents who say “that could be fun” still ensure their “playful” strategy does not appear as blind faith.