Can AI Improve 'Luck', and What Is the True Nature of 'Bad Luck'?

Every time you roll the dice, you get the losing side. In horse racing or the lottery, for some reason, you miss by just one point.

Even though the world should be fair, random biases certainly fall unevenly. These instances of bad luck, which seem like jokes but are not funny, exist as a 'lived experience' that cannot be dismissed as mere statistical error.

Let's try to suppress the urge to lament that we are 'unlucky' and instead understand the structure of that misfortune.

In experiments, unlucky people missed obvious 'accidental hints'.

What is 'luck' in the first place? An experiment conducted by psychologist Richard Wiseman is highly suggestive.

He gave subjects a newspaper and had them count the number of photographs hidden inside. As a result, those who perceived themselves as 'lucky' finished the task overwhelmingly faster than those who did not.

The reason was simple. Lucky people did not focus too much on the task and kept a broad perspective.

On the second page of the newspaper, it was written in large letters: Stop looking. There are 43 photographs in this newspaper. was written in large letters.

People who feel unlucky were so focused on the instruction to 'find the photos' that they missed this obvious 'accidental hint'.

Perhaps people who feel unlucky are simply too serious. They measure the world with too narrow a focus, and the harder they try, the more they push away the accidental opportunities outside their field of vision.

Learning 'failure' more strongly than 'success'

The feeling of misfortune is also related to the mechanisms of the brain.

According to research using fMRI (functional magnetic resonance imaging), people who self-identify as 'unlucky' showed a tendency for specific areas of the brain to overreact to negative results (i.e., misses).

As research by Tali Sharot and others shows, this may mean that the processing of dopamine prediction errors is distorted. Simply put, their brains learn and remember 'failure' more strongly than 'success'.

Even if there are five 'wins' and five 'losses', because the pain of losing is so intense, only the impression that 'there were many losses' is etched into their memory.

Their brains have become devices that process random occurrences as 'personal pain'.

Real-world success depends on past success

It is not just individual issues like perspective or brain response; the social system we live in itself creates 'misfortune'.

Complexity researchers like Pluchino and others conducted an interesting simulation. They had 1,000 virtual individuals with identical abilities and effort compete under the same conditions.

As a result, success was concentrated in a very small group, and at the same time, a 'consistently unlucky minority' always appeared. Why is this? Because real-world success depends on past success (the Matthew effect).

Those who achieve a small success by chance early on are more likely to get the next opportunity.
Conversely, those who experience a small failure by chance early on become more distant from the next opportunity.

Even with the same ability, the accumulation of initial random 'misses' structurally creates a group for whom 'nothing they do works out'.

This is not something intended by anyone, but a cruel characteristic of probabilistic systems.

To tame 'bad luck'

So, what should unlucky people do?

First, lamenting that one is 'unlucky' is, in itself (as Wiseman's experiments show), the entrance to a vicious cycle that narrows one's perspective and invites further bad luck.

What is needed is not lamentation, but a change in strategy.

First, broaden your focus. By shifting your consciousness away from single-minded effort and 'not trying too hard to look,' you will notice hints in your surroundings.

Second, change how you interpret results. Be aware that the brain excessively remembers the pain of 'misses,' and calmly record the fact that 'hits' are also occurring statistically.

Third, understand the system. Recognize that your failures are not necessarily due to your own lack of ability, but may be caused by structural biases in the system.

'Unlucky people' are simply those who have looked more directly at the absurd aspects of the world than others.

Betting is a game of intelligence regarding how to deal with that absurd world.

Do not fear the misses; accept them as data and calmly maintain the rhythm of your next attempts. Let's try changing our perspective, and think deeply about the practice of taming bad luck.

Can AI improve 'luck'—two different technical approaches

The term AI is broad, and 'luck' can be thought of in two separate ways.

One is 'predictive models' represented by machine learning (ML). They learn data patterns in the real world and predict probabilities for the future.

The other is 'interpretive models' represented by large language models (LLM) like ChatGPT. They learn human language patterns and reinterpret the meaning of events.

Both relate to human 'luck,' but their methods of intervention are vastly different.

1. Predictive AI—managing uncertainty and reducing risk

The 'predictive AI' referred to here does not mean mere experiments, but systems that are routinely retrained in the cloud and incorporated into real-world business (market forecasting, user behavior analysis, anomaly detection, etc.).

For this type of AI, 'luck' or 'chance' is uncertainty (noise) that must be managed and minimized.

Its purpose is to extract patterns from vast amounts of data—such as market fluctuations, user behavior, and equipment failure—to predict probability distributions. By doing so, it reduces the impact of 'outliers (i.e., sudden misfortune)' and maximizes expected value.

In short, what this AI aims for is to 'reduce misfortune.'

This is not an occult discussion about 'improving luck,' but rather a strictly rational risk management procedure to build a stable system that is not swayed by luck.

2. Language AI: Rewriting Interpretations and Retraining Cognition

On the other hand, linguistic AIs like LLMs do not intervene in the actual probabilities of reality, but rather in the realm of interpretation, which is how humans perceive 'misfortune'.

As seen in the previous analysis (Wiseman's experiment), the experience of 'bad luck' is deeply tied to psychological factors such as 'narrow-mindedness' and 'excessive focus on failure.' LLMs can help correct these cognitive biases.

For example, suppose a user inputs, 'I failed again. I have bad luck.'

An LLM can ask, 'How would a lucky person interpret that event?' or suggest, 'Can you redefine this failure not as a definitive defeat, but as data for your next attempt?'

This is what is known in psychology as 'cognitive reframing.'

An LLM cannot change the roll of the dice in reality. Instead, it adjusts how our minds perceive the outcome—that is, it adjusts the filter through which we perceive probability.

The realistic support that language AI can provide to 'improve luck' involves preventing us from excessively internalizing misfortune as pain (as seen in the previous fMRI study example) or missing signs of good fortune (as seen in Wiseman's experiment).