Is It Really Better to Be Safe Than Sorry?

“Better safe than sorry” is a well-known and intuitive notion. Yet an intriguing new theory of psychopathology argues that a “better safe than sorry” brain architecture in fact underlies many seemingly unrelated phenomena of psychopathology.

Psychologists have long recognized that many psychological disorders emerge from a personality disposition toward intense negative affect. Current personality theory refers to this dimension as “neuroticism.”

High neuroticism is an across-the-board risk factor for multiple mental health issues. Neuroticism has been shown in research to be a relatively broad and stable tendency, shaped by both genetic and environmental factors. It involves poor emotion regulation and a propensity for appraising situations as threatening, anticipating negative outcomes both personal and global, and experiencing intense negative moods and emotions. People who are high on this trait tend to be more threat-vigilant, behaviorally inhibited, and intolerant of uncertainty and ambiguity.

A recent paper (2020) by Belgian health psychologist Omer Van den Bergh and colleagues has proposed a new theory of neuroticism, viewing it as a product of a “better safe than sorry” (BSTS) brain processing strategy.

The authors begin by taking the view—increasingly accepted among brain researchers—that the human brain is a “predictive processing” system, actively constructing a model of reality by sorting through noisy sensory input using pre-existing, learned information. In this view, “prediction signals from models in the brain are matched with sensory input resulting in prediction errors that are fed back to improve the adaptivity of these models when making perceptual inferences and actively navigating in the environment.”

In other words, the brain works by estimating the likelihood of what new input means given previous experience. Discrepancies between predicted and actual inputs (prediction errors) are corrected in a process of error minimization, until the brain settles on a subsequent model that is most likely to be adaptive.

Several strategies are deployed in this error minimization operation. The brain may adjust prior beliefs to accommodate the input; it may seek to receive more or better input, or it may revise how it evaluates the input. For example, if I’m looking for my truck in a crowded parking lot, my brain generates neural patterns acting as prior beliefs that will facilitate spotting it. If I don’t see my truck, I may decide it was stolen, or I may move to a better vantage point from which to search, or I may decide to focus on my truck’s shape rather than its color to search for it.

In this view, prior beliefs, prediction errors, and subsequent (posterior) beliefs constitute probability estimates of varying levels of precision. For example, if my truck is painted an unusual color (hot pink), both prior beliefs and prediction errors representing color are highly precise, so I will recognize my truck easily. However, in the dark, I’m likely to "recognize" my truck in any other truck on the lot. In this case, my strong prior beliefs dominate the weak input. In daylight, however, I am much less likely to mistake any other truck for mine, because I have access to detailed input that will help me rule out all non-pink sedans.

Our conscious experience is always influenced by our expectations (prior beliefs) to varying degrees, depending on the situation. Highly salient expectations will impact our conclusions greatly if the input is muddled, and vice versa. If I’m sure where I parked my truck, I’ll head to that parking spot. If I’m not sure, I’ll look around the lot for the pink truck.

It is important to understand that in running its prediction and verification processes, our brain is not after maximum accuracy but maximum adaptive utility. Thus, inaccurate beliefs (or those not verified by high-quality input) may be accepted if they are useful for survival. For example, many people have a self-serving bias, seeing themselves as a bit better than they objectively are. This has adaptive benefits as it shores up hope, confidence, courage, and mood—accuracy be damned.

From an evolutionary perspective, neuroticism—a cautious and avoidant tendency—is adaptive to the extent that it protects us from danger. In a world where adversity is common and its consequences dire, the costs of missing a threat are higher than the costs of a false alarm. Thus, it makes sense to bias the detection calculus to privilege looking for, noticing, and responding to threat. In other words: "better safe than sorry."

Alas, this adaptive process goes awry in highly neurotic people as the feedback loop responsible for correcting and updating the preexisting model is terminated prematurely. Specifically, threat processing in highly neurotic persons “can be seen as involving a decision rule that has shifted towards oversimplifying input.” This shift improves the system’s speed in categorizing input as threat by sacrificing the resolution level of the input being processed. It takes a quick, yet blurry, snapshot.

Such low-resolution processing of input means that preexisting threat-related expectations are unlikely to get updated and corrected. “Reduced detail in processing information from the inner and outer world and poor updating of prior beliefs will also result in generative models with low level of detail… leading to chronic conditions of uncertainty/surprise on the longer term.” Thus, “when predictive progress stagnates, the persistent deviations between model-based prior expectations and evidence may engender unproductive coping.” So long as we rely on grainy, low-resolution images, the Loch Ness monster, UFOs, and the Yeti continue to exist.

The authors argue that this “better safe than sorry” strategy allows existing threat-related beliefs to dominate immediate experience, over time leading to chronic anxiety, uncertainty, bafflement, and avoidance. This basic process gives rise to diverse mental health phenomena. For example, perseverative rumination (rehashing abstract worries), deficits in autobiographical memory (inability to recall specific, detailed episodes from one’s past), poor inhibitory fear learning (an inability to learn from experience that something is no longer a threat), and medically unexplained symptoms (symptom reports unrelated to physiological dysfunction) are four separate phenomena independently linked to neuroticism and psychopathology.

The authors argue that all of these phenomena, while differing in content, manifest the same BSTS process, resulting from “applying a perceptual decision rule that has shifted towards oversimplifying input... allowing highly precise threat-related priors to have stronger impact on the eventual categorical danger perception.”

This theory has implications for therapy. If neuroticism indeed reflects a malfunctioning error reduction brain process, then therapy techniques that train clients to face, accept, and tend closely to aversive, threat-related input elements should help repair the error correction process and modify threat-related expectations, resulting in “a more adaptive generative model about the world… characterized by less unresolved mismatch between expected and actual input.”

To keep a clean house, you need to deal with dirt. To overcome fear, you need to face it. To feel less sorry, you need to first accept feeling less safe.

LinkedIn image: Nutlegal Photographer/Shutterstock