Arrangements for
believing oneself

IThe answer in the corner

In the experiment, the correct answer is already on the screen. Participants are asked to judge which of four quadrants contains the most dots. A cue in the lower-right corner can help them answer correctly. They are told that they may look at it, but should try to rely on their own effort. Afterwards, they predict whether they could solve a problem of the same difficulty without the cue.

Dongmei Mei and colleagues varied how conveniently the answer could be read. For one group it appeared as an upright graphic. For another it was written in Chinese and turned upside down. The second arrangement made using the answer more laborious and, the researchers argued, harder to disregard when accounting for a successful response. Both groups could obtain assistance. What differed was the ease with which that assistance could disappear from the explanation.1

The distinction matters more than the familiar observation that people overrate themselves. In the easier-cue condition, participants’ reported efficacy was associated with the researchers’ estimate of prediction error; in the harder-cue condition, measured ability was the significant predictor in their regression analysis. The group-level prediction errors did not simply show universal overconfidence. The authors interpret their findings in terms of ambiguity in attribution, including both overestimation and underestimation. A small assisted success can become material for an inaccurate account of ability without producing the same error in everyone.

This experiment offers a place to begin thinking about “self-psyoping,” a phrase whose military vocabulary makes a rather ordinary arrangement sound like an exceptional conspiracy. A person does something to influence their own condition, then encounters the result as information about themselves. The intervention need not be forgotten. It may remain perfectly available while losing its importance in the account.

Consider, as a separate hypothetical example, a person who maintains that they work best under pressure. They begin late, finish exhausted, and cite the completed task as evidence. The explanation has helped produce a circumstance that subsequently appears to support it. Whether they would have worked better with more time remains untested. The difficulty is not that the claim must be false, but that the person’s procedure for confirming it supplies very little opportunity to find out.

I use recursive self-persuasion for the additional step by which these results modify the account that will organize the next occasion. There is no need to posit a concealed manipulator inside the person. An explanation can do practical work before anyone has decided to make it a strategy.

IIThe internal observer

We commonly give the person who performed an action privileged authority to explain it. They were, after all, present throughout. Research on metacognition complicates this privilege. Performing a task and judging one’s performance are related capacities, but they are not identical. Fleming and Dolan’s review describes evidence that metacognitive accuracy can differ even when task performance is held approximately constant. Confidence must be compared with an independent criterion; its intensity alone tells us little about its accuracy.2

The internal observer is consequently in an awkward position. It has unusually intimate access to the event, but that access does not include an infallible record of the processes that produced it. An account of one’s motives may be sincere and mistaken. Greater attention to oneself may provide more material for an explanation without providing a better means of checking it.

There is also no agreement that every motivated mistake should be called self-deception. Chance and Norton distinguish definitions ranging from a motivated false belief to the much stronger requirement that someone simultaneously hold conflicting conscious and unconscious beliefs. The latter introduces a considerable measurement problem: one must establish both the hidden belief and the sincerity of the expressed one.3 Treating a confident statement as evidence of its own concealed opposite merely relocates the difficulty to the observer.

For this essay, the useful question is therefore less dramatic: under what circumstances does a person’s explanation become difficult for their own experience to correct? Connors and Halligan’s account of belief formation separates the occurrence that prompts a belief from its interpretation, evaluation, acceptance, and subsequent effects on cognition. An accepted belief can influence what is attended to and remembered, supplying material for later belief formation.4 This gives recursion a sequence rather than an aura.

A description such as “independent” can operate at several points in that sequence. It can make an offer of help seem insulting, make refusal seem necessary, and make the resulting isolation seem consistent with character. None of these interpretations requires the others. Together, however, they make the description progressively more expensive to revise: changing it would involve reconsidering the meaning of earlier decisions as well as choosing differently next time.

Stability is not always a mistake. A person who abandons a well-supported self-understanding after every hostile remark would hardly be more accurate. Mokady and Reggev propose that self-concept maintenance can be understood through predictive processing, with an additional hypothesis that confirmation may itself be rewarding.5 Their proposal should not be mistaken for an experimental demonstration of the whole process. It does, however, identify a useful problem: preserving a workable account of oneself and preserving a flattering or familiar account may sometimes employ similar procedures.

IIITwo versions of the image

In their 2013 conference paper Killing in High Definition, Scott Fitzsimmons and Karina Sangha consider the psychological strain of operating remotely piloted aircraft. They reject the assumption that physical distance necessarily produces emotional distance. Operators may watch people for extended periods before an attack and see its consequences afterwards. They also describe the abrupt movement between a classified working environment and domestic life at the end of a shift.6

The paper becomes particularly disturbing when it turns from explanation to design. The authors recommend overlays that would appear when a weapon was launched and replace the visible human victims with simplified graphics. Commanding officers would retain access to an unmasked feed; the unaltered images would also be archived. The event would remain available to the organization while being made less perceptually available to the person carrying it out.

“these overlays are not intended to make it easier for operators to kill.”

Fitzsimmons and Sangha, Killing in High Definition, p. 11.

The distinction the authors draw is between facilitating an act and helping someone cope with having performed it. That distinction describes the stated purpose of the proposal, but it cannot establish the proposal’s effects. The fact that attacks already occur does not show that changing what operators see would leave judgment, attention, or later conduct unchanged. Nor does retaining a full recording elsewhere settle what it means to remove information from the immediate encounter.

The authors’ argument is directed towards the welfare of operators. It draws on military reports, journalism, and earlier scholarship rather than testing the proposed interface. Reading it critically does not require dismissing the distress it discusses. It requires noticing that the remedy redistributes access to the consequences of an action. One part of the institution continues to know in a form another part is no longer required to see.

Ordinary self-description is not equivalent to warfare. The relevance here is a narrower matter of mediation. An account of an event can preserve propositional knowledge while reducing the prominence of particular details. A person can know that help was received and nevertheless experience the outcome primarily as an achievement of their own ability. They can know that a record has been selected and still use its apparent consistency to judge themselves. Knowing how a representation was made does not automatically neutralize its effects.

The grotesque feature of the proposed display is its administrative ordinariness. There are separate views for separate responsibilities, and a complete record in storage. Nothing needs to vanish from the system as a whole. It need only be absent from the place where it would interrupt the next action.

IVThe second explanation

A first-order justification explains an action: I refused assistance because I am independent. A second-order justification explains why objections to that account should not be considered: other people want me to doubt my independence. At this point, criticism has acquired a role inside the explanation it was intended to challenge. The more criticism arrives, the more occasions there are to rehearse the account.

This is where the vocabulary of self-psyoping becomes interesting and hazardous. Recognizing an influence attempt can itself organize resistance. Fransen, Smit, and Verlegh distinguish several ways of resisting persuasion, including avoiding a message, contesting its source, selectively processing its content, and reinforcing an existing attitude. Concern about being deceived is one of the motives in their proposed framework.7 Suspicion can help someone examine a claim, but it can also furnish a reason to leave their own claim unexamined.

Applied inward, the same difficulty produces an indefinite series of inspections. A conviction is suspected of serving a motive. The suspicion is then suspected of serving another motive. Each interpretation seems more sophisticated because it occupies a position above the previous one, although no new evidence may have entered the discussion. The person is doing more interpretive work while becoming less able to decide what would settle the matter.

It would be easy to make this essay reproduce that arrangement. It could redescribe disagreement as defensive self-persuasion and take assent as confirmation. The resulting theory would accommodate every response by denying that any response could count against it. Calling that procedure recursive would not improve it.

The more modest alternative is to examine specific arrangements. What was recorded? What was omitted? Which outcome was predicted in advance? Did the explanation change after the outcome became known? These questions do not require a final verdict about whether someone is fundamentally honest with themselves. They concern the provenance of the evidence and the rules used to interpret it.

Such arrangements are rarely wholly private. A performance review provides categories in which a worker must become intelligible. A platform provides counts that make some forms of activity easy to compare and others difficult to describe. People can adopt these measures for their own purposes, but their apparent self-authorship does not erase the institution that supplied the form. The question of who is persuading whom cannot always be answered by looking for a single intention.

VConditions of revision

There is no position outside influence from which a person can compose an entirely unconditioned account of themselves. Reminders, commitments, rehearsal, and deliberate changes of environment are ordinary means of making action possible. Their value cannot be decided merely by observing that they alter belief or conduct. The important distinction is whether the account can still be corrected, and what correction would cost.

Outside feedback offers one possibility, though not an automatic solution. Bollich, Johannet, and Vazire distinguish actual information from other people from our imagined account of what others think. The latter can be generated without encountering anything new. Their review also notes that knowledgeable others may have biases of their own, and that credible feedback can be difficult to accept.8 Independence of evidence is a practical problem, not something guaranteed by the presence of a second person.

In the dot-discrimination experiment, the source of assistance occupies a visible position on the screen. Outside the laboratory, assistance and constraint are harder to isolate. A finished task contains other people’s labor, inherited skill, favorable circumstances, and costs that may not become apparent until later. Any description necessarily leaves something out. The issue is whether those exclusions remain available for reconsideration or become prerequisites for continuing to recognize oneself.

The relevant freedom may be quite unheroic: to record that an earlier judgment was wrong without converting the correction into a new verdict on character. An account of oneself can remain useful while ceasing to function as an instruction for every subsequent occasion. The completed task would then remain completed. It would no longer have to prove who completed it.