The Geometry of Illusion — Part IV

There is a difference between wanting an answer and wanting a particular answer. It sounds elementary until the answer concerns ourselves. Then the two desires become remarkably skilled at impersonating each other.

Part I asked whether passions could produce recognizable classes of interpretive error. Part III used Evagrios to separate an object from the attachment formed around it. Now we need a more exact account of what a preference can do to an argument. The research term is motivated reasoning: the influence of goals or desired conclusions on the way evidence is accessed, constructed, and evaluated. Kunda’s influential review distinguishes a goal of accuracy from a goal of reaching a directional conclusion. The latter does not grant unlimited freedom to believe anything one likes; a desired conclusion still needs a justification that seems plausible to the thinker. [1]

That constraint makes the mechanism interesting. A person seldom says, “I prefer this verdict, so I will ignore the case.” He says, “I have examined the case, and fortunately the evidence agrees with me.”

Imagine receiving a short message from someone whose approval matters to you: “Fine. Do what you think is best.” You suspect anger. You reread the punctuation, the delay before the reply, the missing emoji. Each detail is real. But the evidential standard moves with the desired conclusion. A delay is decisive when it supports anger; the sender’s known habit of answering late is dismissed as an exception. If the message had arrived instantly, perhaps the speed itself would have seemed cold. A theory that can recruit every possible observation has become difficult to falsify. It is also an excellent employee of the ego.

In statistical terms, evidence should update competing hypotheses according to how likely it would be under each. For hypotheses H₁ (“the sender is angry”) and H₀ (“the sender is not angry”), an observation E is informative through a likelihood ratio:

Λ(E) = P(E | H₁) / P(E | H₀).

If delayed replies are common under both hypotheses, the delay is weak evidence. If I count it as strong only when I need anger to be true, the defect lies in my weighting of evidence, not in the message’s existence. This equation is a framework for reasoning, not a claim that we know the relevant probabilities in an ordinary conversation.

Raymond Nickerson’s review describes confirmation bias as selective seeking or interpretation of material favorable to a belief or hypothesis. [2] The term covers several processes; it should not become a diagnosis delivered whenever another person disagrees with us. Otherwise the concept performs the very bias it names: every objection becomes evidence that the objector is biased.

The boundary between reasoning and seeing also matters. Balcetis and Dunning reported experiments in which motivational states affected responses to ambiguous visual figures. [3] But the interpretation of “top-down” effects on visual experience is contested. Firestone and Scholl argue that attention, memory, judgment, and response can produce findings that look like changes in perception without establishing that cognition altered visual processing itself. [4] One need not settle that debate to make the present point. Our message example concerns an inference about intention, not the optics of the eye.

Part I’s formula P = T_S(R) becomes more useful if we do not force all errors into P. A more careful sequence would be:

event → selected evidence → interpretation → expressed judgment.

A motive could act at any stage. It might draw attention to the punctuation. It might influence the meaning assigned to the delay. It might leave interpretation intact but alter what the person is willing to say. Each possibility calls for a different test. The word distortion names a problem; it does not locate the mechanism.

What can a person do before the verdict hardens? Write down the rival explanations. State what observation would favor each. If a reply tomorrow could count as proof of both anger and affection, revise the test before tomorrow arrives. Ask whether the same evidential rule would be applied if the desired conclusion were reversed. These are not rituals of purity. They are ways of making a private standard visible enough to challenge.

The unsettling conclusion is that a mind can build a valid-looking argument from genuine facts and still be answering a question it secretly chose in advance. The facts are not false. The proof has been given the wrong job.

Notes on evidence

The message and likelihood ratio are illustrative. No probability is estimated for that example. The cited studies concern defined experimental or reviewed research settings; they do not establish a general equation connecting passions to visual perception.

1. Ziva Kunda, “The Case for Motivated Reasoning,” Psychological Bulletin 108 (1990), 480–498.
2. Raymond S. Nickerson, “Confirmation Bias: A Ubiquitous Phenomenon in Many Guises,” Review of General Psychology 2 (1998), 175–220.
3. Emily Balcetis and David Dunning, “See What You Want to See: Motivational Influences on Visual Perception,” Journal of Personality and Social Psychology 91 (2006), 612–625.
4. Chaz Firestone and Brian J. Scholl, “Cognition Does Not Affect Perception: Evaluating the Evidence for ‘Top-Down’ Effects,” Behavioral and Brain Sciences 39 (2016), e229.

Previously in The Geometry of Illusion: Part III.