Generalisation is difficult because it requires you to find something in common with all members of a group. People often make false claims about everyone in a group by taking what only part of that group do or are, and then claiming that everyone in the group do or are that.
A common event is to take a small group as representative of the larger group without determining a reason why. If someone has five friends from Japan who all have trouble saying words with the letter “R” in English, then it is natural to suspect that the Japanese language does not use the letter “R” in the same way as we do, or that the Japanese are incapable of saying it correctly in English. It is fine to suspect this, but it is incorrect to assume that it is true without proper evidence of causality.
A form of generalisation is induction, a process used in math where a general rule can be determined from particular instances. The mathematical processes to perform induction require you to prove beyond any reasonable doubt that the specific instances necessarily have a relationship to a broader rule. While this level of work is inconvenient or even impossible for the humanities, one should at least attempt to do so instead of simply assuming that individual actions imply something about a whole group.
If every African American you know is a thief, then it is easy (but lazy) to say that all African Americans are thieves. This might be true if there were a valid reason to think that the ten, or hundred, or thousand African Americans you know could represent all African Americans. Such a reason is rare and very difficult to find. If we say that theft is genetic, then we have to prove that all African Americans have this gene. If we say that they have all learned it, we have to show how it is even possible for every single African American to learn the exact same thing, without any exceptions. If there are any exceptions, then they have to be justified, otherwise they might prove that the rule isn’t a rule. The rule might just be a coincidence.
Commonality does not imply Causality
Even if we prove that all currently living African Americans were thieves, we need to show that this is a causal relationship, not just a commonality. An example of why that matters: If I were to invent a new type of cheese, but I made a mistake in the production, and the first ten wheels of this cheese comes out poisonous, then we can say that every wheel of this cheese is poisonous. In this example, every single wheel of this type of cheese is poisonous, so they all have this in common. However, since this was a mistake, it doesn’t mean that this type of cheese is inherently poisonous. As long as the production mistake is corrected, it is likely that the next wheel of cheese will not be poisonous. This shows that even if everything or everyone in a group share something, that does not mean that people who join the group later will also share it, or that previous members of the group shared it. Even if every living African American on the planet were a thief, that does not necessarily mean that African Americans are inherently thieves, or that every African American in history was a thief, or that all African Americans who are born in the future will be thieves. The key here is that we said all -living- African Americans. We could also say “all living and dead African Americans are thieves”, but that wouldn’t mean that future African Americans would also be thieves. When we talk about a specific group with a specific feature, it is usually limited to exactly what we included or excluded. If we talk about living African Americans then it usually a conclusion that only applies to -living- -African Americans-. By defining the group we have limited what we can say about a different group, such as future African Americans who do not necessarily share traits with African Americans of other time periods. To make this generalisation we need to provide a causal relationship, otherwise we are just making assumptions.
Statistics and Sample Size
If a causal relationship is too difficult to determine, then we cannot know if a generalisation is true. We can approximate it, if need be, using statistics to determine how large a sample size we need for it to be accurately representative of the whole group. If someone lives in a country with very few foreigners, and meets five different people from the same foreign country who all share a feature, it is normal for that person to think that it would be a huge coincidence if it were only these five people. Instead, it sounds far more likely that the reason these five people (who might not even know each other) all share this feature is because they all share being from the same country. While this may be possible, further evidence would be required. Without that evidence, statistics would be the best way to make an informed guess, but five people is not a large enough sample size. Even with 100 or 1000 people, it would not make sense to claim that they could represent a million people from the same country. The mathematics show that such a relationship is unreliable.
Despite a lack of causal relationship, a large enough sample size, or real evidence, people often still insist that they know. This might be simply because they are very confident, or because they trust in their instinct which tells them this. This is textbook closed mindedness and denial of reality. Instinct is not infallible, and insisting that you are right or “trust me, I know” isn’t how reality works. You can be as confident as you want about your assumptions, but that doesn’t make them any more accurate.
What can we say about a whole group then?
There are certain features that can be said about an entire group. These are usually trivial things. They are the very features that define the group. We can safely say that all members of the set of African Americans are African and American, or we can say that all prime numbers are prime and numbers. While this may sound useless, it is the first step in properly defining generalisations. If we take any group, we can make a list of things that are true about everything in the group. This list would start with the defining features of the group which are taken as obvious, but they may lead to other conclusions.
When we want to make a generalisation, we must use inductive logic or scientific method to do so.
Science works by analysing data collected from a specific experiment, and trying to figure out whether or not something meaningful can be determined from the data or not. To do this they try to collect as many examples as possible. This is done to make sure their sample size is large enough, reducing the chances of exceptions or coincidences affecting their conclusions. They might conduct an experiment thousands of times so they have more than enough data that fit a pattern reliably. Scientific phenomenon often act identically as long as you can isolate them, but groups of humans are harder to determine because we have free will and have different genetics and upbringings. All electrons share very specific qualities, and we classify them as electrons because of their fundamental nature.
Groups are often arbitrarily classified
In fact, the groupings we make for people are often arbitrary. African American isn’t really a race so much as a common heritage, which is only one of the many variables with which we could group people. We could instead group them by dark skin colour, which would include many people who are not from Africa. If we grouped people from Africa, that wouldn’t even be just African Americans or dark skinned people, it would include many people with white skin as well as Muslims. There isn’t even really a proper definition for African American, because while there may be agreed upon definitions, they don’t have the same genes or really any meaningful shared qualities. They are as diverse as any other race, because races are arbitrary groupings of extremely diverse people with a very vague common skin colour or whose ancestors lives in a certain place.
If we split people up into different groups based on arbitrary things, then it doesn’t make any sense to expect them to share many qualities.
Some people claim that you should “never hit a woman”. I would say that you shouldn’t hit anyone, regardless of their gender. Such people might say that women particularly should not be attacked because they are physically weaker than men and so it’s not fair because they cannot defend themselves. This is an unfair generalisation of women, but one could argue that on average most women are weaker than men, if only by statistics. If that is true, then not hurting women would mean that you are mostly not hurting people who wouldn’t be able to take it. If your goal was to avoid hitting people who cannot take it, then this would be consistent with their logic. By not hitting women you’ve avoided hurting a lot of defenseless people. The problem is that their expression “never hit a woman” doesn’t account for weak men. If they follow this expression to avoid hurting weak people, then they run the risk of hurting weak men. This expression is clearly insufficient to meet the needs of their goal. They will avoid hitting strong women because they’ve assumed they’re weak, and will hit weak men because they assume they’re strong. Their logic is inconsistent and flawed because they generalised incorrectly.
This example shows how something that sounds approximately right can end up not working. The logical error can be seen better in the form of a syllogism:
- A. Weak people should not be hurt
- B. 90% of women are weak
- C. Women should not be hurt
For this to actually work, B should instead say “All women are weak people”. As it stands now, it is a false syllogism because the conclusion does not follow from the premises. Some might argue that 90% is good enough though, because if we protect women then we protect a lot of weak people, even if not all of them are weak. This is a fallacy that many people use to further their own agendas using faulty logic. Instead of saying that we should not hurt weak people, they pick a specific group of people who are mostly weak. There is no advantage to picking such a group, it only has the disadvantage of being less accurate. It is less accurate because we are skipping a step, making the logic indirect. Instead of directly applying your logic to the group that needs it, you are skipping them and indirectly applying it to one of the subgroups.
All it takes to correct this is to say “never hit someone who can’t take it”. This would include weak women, weak men, or anyone else. This now covers every group of people who should not be hurt, so it works out much better than just saying women should not be hurt. If your reason for not hitting someone is because they are weak, then don’t be indirect and look for a group that is mostly weak, instead just don’t hurt weak people. If your reason for not hurting someone is because they can’t fight back, then “don’t hurt people who can’t fight back”. It’s that simple.
Choosing a group that mostly fits this makes no sense and betrays your own goal.
Generalisation is useful when done properly, but otherwise is harmful and makes assumptions that limit our thinking.