How To Maximize Learning From Other People’s Mistakes and Failures

Your own mistakes and failures are limited, so leverage other people’s experiences

While striving in our lives working towards success, it is inevitable that mistakes and failures will occur. To deal with them, there exist aphorisms about the proper mindset to take these mistakes and failures in stride so you can improve in a specific endeavor and in order to grow as a person.

While this is 100% true, it may be more beneficial to learn from the mistakes and failures of others.

I will consider here what types of feedback and learning can result from mistakes and failures, and how this process can be made more efficient by consider other people.

Feedback During Learning

Types of Feedback

In general, learning requires multiple trials that may involve success and failure, with each trial representing an opportunity to learn and improve. There are different types of feedback of varying frequency and quality.

• Outcome Feedback
  This is the most coarse-grained and most frequent type of feedback and only tells you how well you did at a task.
• Informational Feedback
  This type of feedback is of higher quality since it not only tells you how well you did, but also provides an additional level of granularity as to what you did wrong.
• Corrective Feedback
  This is the highest quality but the most infrequent type of feedback since it also goes into the next step of telling you exactly how to fix it any mistakes or failures.

In an ideal world, we get immediate, specific, and actionable feedback that we know how to use to learn. Unfortunately, learning tasks in the real world may give us very amorphous and unclear feedback that may not just fail to teach us, but may actually teach us the wrong things.

Let’s consider these three types of feedback in a specific case of a beginner learning how to play poker, which is a cleaner and more sanitized version of the real world. The examples I use will be drawn from No Limit Texas Hold’em Poker, which is one of those popular variants of poker.

Outcome Feedback

When a beginner starts to play poker, they may lose or win money for the entire session. This win or loss provides feedback, but it may be low-quality outcome feedback since much of it can still be subject to the wrong interpretation and lessons.

The beginner may play poker correctly and still lose money, and therefore learn the wrong lessons going forward.

• For example, they may only wait until they have premium starting hands like Aces or Kings, and correctly play them by betting bigger.
• But they may still lose money if they bet money but yet other players still draw out against those premium hands.
• Thus because of the penalty of losing money, the beginner may learn the wrong lessons of not betting big and aggressively with those premium hands. Specifically, the pain of money loss reinforces a mistaken outlook that betting big with hands may just lead to losing money.

Perhaps even more dangerous is that they may play incorrectly and win money, and also learn the wrong lessons going forward to set them up to apply those lessons when much larger money is risked.

• For example, the beginner may enjoy playing any trash starting hands without regard to their strength, or table position, or pot size, or any number of other factors.
• But yet despite playing incorrectly, they may still win big just due to pure luck, especially since no starting hands are too significantly disfavored versus other hands in the short term, and it is only in the long term that the probabilities work themselves out.
• Thus in the future, the beginner may learn that play bad starting hands are actually profitable, based on the outcome information they received.

Both of these situations of losing despite playing correctly or winning despite playing incorrectly may lead to low-quality learning since there’s not much information or incorrectly interpreted information in pure outcome feedback. It is only in the limit of a large number of trials that outcome feedback may be useful.

In many ways, biological evolution is an example of outcome feedback since the evolution of a species is subject to the outcomes of each individual member for that species experiencing outcome feedback. It benefits at the aggregate level over many trials, even though it is not very useful at the individual level.

Informational Feedback

This type of feedback is a bit more granular and would involve knowing exactly where that beginner is making incorrect decisions while playing, but without any specific steps to correct it.

This might involve reviewing and scrutinizing a list of all the hands played during the night. From that list exact decisions leading to specific wins or losses at the level of a single hand can be carefully scrutinized to figure out why money was won or lost.

Since complete and detailed information of online poker hands can be downloaded, this type of feedback is available for research.

• For example, this might show that when premium starting hands like Aces are played, they usually win money
• But it may point to the baffling scenario where those same Aces also tend to be big money losers

This informational feedback may be useful to find out exactly where problems arise, they may not tell how to solve them. It’s more useful and actionable than outcome feedback since it’s more precise.

Corrective Feedback

Finally, this corrective feedback involves a step further than simply identifying problems and provide prescriptive actions to fix it. It is usually given by an experienced teacher, instructor, or coach and may involve more abstract poker theory or probability and statistics.

Taking the example from the previous informational feedback for losing money with the Aces, it might involve the following correctivfe feedback.

• Whenever you have Aces, you have to bet even bigger in general.
• This may mean to re-raise bets in order to limit the field of opponents you face and to cut down on the implied odds to make it mathematically incorrect for your opponents to draw more cards on you, so you maximize your chances of winning the particular hand while also committing more capital whenever you have a larger edge.
• This may also go into more theory about how your chances of winning against multiple opponents in the aggregate are much less than when you can isolate one of them, or even about the broader aspects in the theory of poker and implied odds in the abstract.

Although corrective feedback is the best for learning, it is scarce source of feedback since it may require highly-detailed, contextual, and tailore expert analysis and appropriate corrective actions. The quality of corrective feedback must be balanced by the greater availability and higher frequency of outcome feedback or informational feedback.

Thus, it’s important to distinguish these three types of feedback when considering what you can learn from the mistakes and failures of other people, as well as in analyzing your own mistakes. Let us now turn to how mistakes and failures in ours and other people’s lives can be used.

Personal Experiences

The Utility of Mistakes and Failures

Lack of mistakes and failure means that you aren’t trying hard enough to push the boundaries in your learning or challenging yourself.

So long as they are not too devastating, mistakes and failure may serve the purpose of providing very valuable feedback that can aid in the learning process.

Character Building

Of course, mistakes and failures may go beyond simply pedagogical reasons and start to involve building character. We all know of stories of how hardships and from mistakes and failures can build important character traits.

Time is Limited

Even if we are fully aware of the utility of mistakes and failure and fully embrace them or actively seek them out as a useful process of improvement, we still run into a big problem. That problem involves not having access to enough trials to make enough mistakes or fail enough.

Furthermore, some activities such as flying a plane necessarily require that mistakes and failures are relatively small in magnitude.

For example, it may be useful to learn through experience that running out of fuel while piloting a plane is fatal. But the price of this mistake may be too high on an absolute level in terms of the outcome itself, or on a relative level with respect to the lessons that can be learned.

Therefore when time is limited, it is crucial that we learn from other people’s mistakes and be able to leverage the experiences of others.

Leverage Other People’s Experiences

Only Success Tends to Be Celebrated

The problem with the accounts of the success of people is that only the highly visible successes and victories are recounted at length in autobiographies and various documentaries.

This makes sense since these successes may serve to boost the egos of those who document them and may be inspirational to those who learn about them. No one wants to learn too much about all the myriad failures and how bad they felt. They want to know the end success as something to aspire to.

However, that success might have involved the following.

• Success was only one outcome of very many
• Came as a result of sheer luck without any specific actions that warranted the success
• There are more and better lessons derived from a higher number of mistakes or failures

Mapping Out Mistakes and Failures

Thus, it’s important to completely map out the mistakes and failures of other people, and not be too enamored by success.

If you can get a broad set of examples of these mistakes, it is equivalent to your living multiple lives and learning in a much broader way across more situations than if you were to commit all these mistakes yourselves.

Furthermore, learning from other people’s mistakes does not come with it a personal emotional toll. Even if we know that mistakes and failures are useful doesn’t necessarily mean they are fun to go through. At least learning from all the mistakes that are possible from others doesn’t come with that toll.

Finally, it may be that when you get feedback from your own mistakes and failures, it may be that they are of lower quality. For example, you may only get outcome feedback, while studying another person’s life may involve informational feedback, or even corrective feedback if there’s commentary by a more expert analysis. This expert analysis could even be performed by the person experiencing the mistakes.

When Mistakes are Not Really Mistakes

It’s interesting to consider mistakes from a Machine Learning (ML) or Reinforcement Learning (RL) context in particular or an Artificial Intelligence (AI) context in general.

In these contexts, what constitutes a mistake can actually be semantic in nature only. Why?

It’s because “mistakes” in the form of negative examples are crucial and are just as important as the “success” of the correct answer in training a model. In fact, mapping out the space of “failure” in terms of negative examples may be even more important.

In order to distinguish correct model building in ML or RL, it is crucial that the broad space of negative examples are understood as much as possible.

In many ways, ML or RL is the clean and formal version of learning from mistakes that were “committed” elsewhere and not as a result of this specific training.

Summary

• Although learning from your own mistakes is useful, it’s even s more useful to leverage other people’s mistakes so you can improve faster and more broadly with a larger reservoir of examples.
• Feedback for mistakes or failure can be of three different kinds: outcome, informational, or corrective.
• Mistakes or failures derived from other people’s experiences may help you effectively increase the number of mistakes and failures you make without suffering the consequences.