The false alarm trap

Avoiding scary things can keep you safe, but doing it too much is a prison.

It’s probably bad when the inside of your mind looks too much like this. Source

In the last few years, a very interesting conversation has been developing about the personality trait negative emotionality — also known as dispositional negativity, or more commonly, neuroticism.¹ To jump into this conversation, start with the Psychology Today article, Is It Really Better to Be Safe Than Sorry?, by Noam Shpancer, which is itself a gloss on a paper by Van den Bergh et al. (henceforth VDB) that Scott Alexander got excited about back in 2021. For this post, the PT article and a quick skim of the pictures in Scott’s post are probably sufficient background.

Shpancer, Van den Bergh, and Scott all want to know one thing: what is up with negative emotionality? Why do people have this trait? Why does it seem to cause problems for some people? How can these people be helped?

Here is my synthesis of the conversation so far:

• Negative emotions are just another part of our cognitive architecture, which is trying to model the world so that we can act in accordance with our interests. Rationality helps you win, i.e. survive/reproduce/enhance your evolutionary fitness, so it’s to be expected that the principles of rationality have in some way shaped our minds… even at deep, subconscious, emotional levels that we sometimes struggle to exert deliberate, conscious, rational control over.

• Negative emotions like fear, anxiety, shame, embarrassment and disgust² push us to remember threats — conditions can harm us in the physical and social sphere — and avoid them. So when we talk about negative emotionality, we are talking about the problem of how to deal with threats.

• Some threats are severe or catastrophic. Things like big carnivores, poisonous plants, and social estrangement present such a high risk to survival that natural selection will favor people who have a severe threat response mechanism. These people are mentally and emotionally primed to classify certain things as “never get near this / never let this happen” type things and act accordingly.

• There is a dark side of having a severe threat response mechanism. You’re modeling and classifying threats using statistical, experiential learning based on imperfect information. What happens if you misidentify a condition as a severe threat, when it’s really not? First, your evidence threshold for reversing your judgment becomes very high. Since you’ve identified the threat as severe, you’ll require an extreme level of evidence to reverse your judgment. To make matters worse, learning about a thing often involves some sort of contact with the thing. Contact is exactly what you’re NOT supposed to have with severe threats! So to fix your mistake, you must somehow get a ton of evidence about something without getting near it, which seems impossible.

This last bullet point is my own contribution, synthesizing the insights of van den Bergh et al and Scott Alexander. VDB fingers “threat-related categorical³ priors” as the key actor in what goes wrong in the minds of the dispositionally negative, while Scott characterized it as the “trapped prior” problem — the key word here being trap.

The central idea this conversation is that dispositionally negative people tend to see things as more threatening than average. In very hostile environments, this threat-vigilant disposition might help them to avoid (and help others avoid) severe threats better, but it also means they are more prone to become trapped in severe threat misidentifications that are nearly impossible to reverse.

Let’s make up some new terminology to help us remember what we’ve learned here. We’ve got this problem that the mere act of living forces us to solve, the problem of avoiding severe threats. We solve it using a statistical predictive framework with a certain false positive rate. Let’s call the prediction of a severe threat an alarm. Then the problem of mispredicting a severe threat could be called the false alarm unlearning problem, or (more pessimistically) the false alarm trap.

The false alarm trap in insurance

This is not just a psychological problem for individual people to struggle with in their minds. As the business of managing risk, insurance is also subject to a false alarm trap, and I think it validates the theory somewhat to note how universal the mechanism can be.

Insurance underwriting is very complex, and this is necessary for the survival of insurance companies. For example, in life insurance, some people die early because they engage in dangerous hobbies like skydiving. Say something like this becomes known to one company, and they start charging extra or even declining applications from such individuals. Soon the practice will spread, because those riskier people will flock to any company that doesn’t guard itself, and that company could see anything from an erosion of its profits to a complete solvency catastrophe.

Because of market dynamics like this, companies are under constant pressure to measure risk accurately, even though

• The risks are complex, e.g. endless medical conditions that modulate mortality risk

• The risks are changing, e.g. due to medical advances or new environmental hazards

• Information about the risks is changing, e.g. the health information accessible to insurers changes over time, and it’s constrained by the desire of applicants for a streamlined convenient process

• There’s a whole lot of uncertainty. I mean how much risk does skydiving add to a 40 year old male? How do you even figure that out? Where’s the data?

With such a complex, risky, uncertain, and occasionally urgent environment, it makes sense that insurers keep themselves safe by having some hard-and-fast rules about who they can insure. But every such hard rule means the company doesn’t get data about the type of person they turned away. That data is precious! The easiest way to know the nature of a risk is to have lots of direct experience with that risk. So rules are hard to dislodge because they hinder collection of new data, and without new data, all you can say might be “this rule says this risk is unacceptable to us”, year after year.

An insurer in this condition has fallen victim to the false alarm trap, and may see their market position erode as customers prefer competitors with more relaxed risk postures and less hangups about things that don’t really matter that much.

Some solutions

Returning to the psychological realm now… some of the more influential clinical psychology theories, such as cognitive-behavioral therapy and acceptance & commitment therapy, can be seen as focusing on different aspects of the false alarm trap.

Cognitive-behavioral therapy (CBT): many of the cognitive distortions that CBT aims to challenge — all-or-nothing thinking, labeling, overgeneralization, discounting the positive, emotional reasoning — can also be seen as highly effective mechanisms for rapid threat detection and response in a complex and dangerous environment. In a person who is overwhelmed by anxiety, it’s likely that these mechanisms are overactive, and CBT suggests that the patient use higher-order cognitive processes to calm these mechanisms down. CBT works on the first part of the false alarm trap, the part where our standard of evidence to reconsider triggering our alarm is impossible to meet.

Acceptance and commitment therapy (ACT): while CBT revolutionized clinical psychology, ACT challenged the idea that all psychological problems can be reduced to developing rational responses to cognitive distortions. First, some of the causes of our negative emotions might be deeply rooted in the subconscious and hard to challenge, and we might be better off accepting that these negative emotions are okay to have and not catastrophic, even if we hope to someday lessen them. The process of CBT-style rational challenge means spending more time thinking about the threat, which can make it loom even larger than it already is. Second, a lot of people try to make their negative feelings go away by avoiding the situations that trigger them, but avoidance can make things worse, so sometimes it’s best to “suck it up” and face what we’re afraid of. ACT works on the second part of the false alarm trap, the part where the alarm prevents us from gathering new evidence.

Conclusion

That’s all for now. I’m surprised I ended up writing this much.

I’m very interested in this topic and look forward to any thoughts you may have.

1. I prefer negative emotionality to neuroticism because the label of neuroticism reduces a major personality trait to nothing more than “a tendency to have neuroses [mental health problems]”, which reinforces the pathological fixation that modern psychology has been subject to since Freud.

A focus on pathology is not bad per se — a great deal of medicine and technology and human endeavor is about solving problems — but it may lead to a distorted picture of the human person and human condition. It is much easier to understand malfunctioning machinery if we know how the machinery is supposed to function.

2. Put aside sadness for the moment, it does something else.

3. This is very close in spirit to my social norms paper, which also sees a black-and-white, good-or-bad, categorical type approach as necessary for practical response to threats from the behavior of our fellow humans.