In many Youtube videos & comments (regarding such issues as the new Youtube Terms Of Service or Milo Yiannopolus being banned from Twitter) I have noticed people using a particular form of fallacious argument, which I shall refer to as The 'Inconsistent Application' Fallacy.
This fallacy has a long history in other settings. It has frequently been used by groups of people with certain political axes to grind to accuse the police and legal system of favouring one group of people (e.g - a racial or religious group) above another, and 'inconsistently' applying the laws on that basis.
In all cases where an 'Authority' (whether a legal system, a social media outlet, or whatever) is accused of inconsistent application of their rules to favour one group over another, one should never rule out the possibility that the accusation might be correct. However, proving that conclusively would require the correct standard of evidence, and unfortunately too often the accusation is merely based on selective data (a few cherry-picked examples), confirmation bias, and lack of overall statistical data (in order to judge if the examples given are representative of the whole)
This fallacy comes in a two slightly different forms, and I shall examine these separately
(FORM 1) Person from Group A does X - The 'Authority' punishes them.
Person from Group B does X - The 'Authority' doesn't punish them.
Therefore the 'Authority' favours Group B above A.
Possible Reasons Why That Conclusion May Be False ~
(one) Lack Of Overall Statistics
In the case of social media and internet sites such as Youtube, there are millions of users. Any single user can only ever be aware of the tiniest fraction of fellow users & what these users experience. What is experienced by the vast majority of other users is totally 'invisible' to us. We cannot draw an accurate conclusion regarding the overall picture based on the tiny 'sample group' that any of us can only ever have access to do. We may, in our own little tiny corner of an internet site, have seen three people from Group A have their accounts terminated & nobody from Group B, but how can we possibly know whether or not that is representative of the total statistics ? (perhaps overall, actually more people from Group B are having their accounts terminated ?) We can't possibly know the answers to these questions (& to claim otherwise would be committing the Fallacy Of Omniscience).
Internet sites do not publish any statistics that could be used to draw accurate conclusions regarding these questions, so speculating a 'conclusion' can only ever be indulging in Confirmation Bias, and should be dismissed as such.
In the case of a legal system (the courts & the police) overall statistics can be obtained. However, too often when people are making the Inconsistent Application Fallacy, it is not based on the overall statistics. Instead the argument is just based on two cherry-picked stories from this week's newspapers (e.g - in one case, a white person committed a crime & got punished, and in another case, a non-white person committed the 'same' crime and didn't get punished) and may not be representative of the overall statistics.
(two) Is 'X' Objectively Quantifiable ?
On the internet, the argument that people from Group A are "being punished for X", whilst people from Group A are "not being punished for X" is almost impossible to present as an objective argument if the 'X' in question is not something that can be objectively quantified in any absolute sense. For instance, people might have their accounts on a particular site terminated for 'Hate Speech', but most internet sites cannot provide an absolute definition of that term that could be universally applied to every possible interaction on that site (there's just too many potential variables).
Therefore, if one person gets their account terminated for 'Hate Speech', any individual can argue afterwards that the decision was 'right' or 'wrong'. Equally, if a person thinks a particular user is guilty of 'Hate Speech', reports them, and the site administrators decide that user is not guilty, any individual can argue afterwards that the decision was 'right' or 'wrong'.
However, in the lack of any absolute objective criteria for 'Hate Speech', all such arguments are likely to be based on nothing more than subjective & biased assumptions.
(three) Problems With Administration / Inbuilt (Non-Ideological) Inconsistency
Many internet sites are notoriously badly administered & inconsistent with decision making (regarding people reporting & flagging certain postings). Their reporting system is not run by the rigid standards of a court of law, and instead there appears to be a vast group of inadequately-trained administrators, who are not working from clear & definite guidelines, and are often making completely arbitrary decisions because of all that. Decisions are often very inconsistent , but this has nothing to do with them favouring one group of people over another, but instead has everything to do with all these random factors, such as which particular administrator is dealing with a particular report, and how they have chosen to interpret the very loose guidelines they have been given.
In any given period of time, some decisions are going to look like they favour one particular group, but equally, other decisions are going to look another particular group. In these cases, I wouldn't even necessarily suspect that it's individual administrators allowing their own ideological bias to affect the decision. Instead, I think it's just the inconsistency that's inevitable in such a badly designed reporting system.
(FORM 2) Person from Group A commits crime X.
Person from Group B commits crime X.
Person from Group A gets lesser sentence than person from Group B.
Therefore Group A is being favoured over Group B.
Possible Reasons Why That Conclusion May Be False ~
(one) Overall Statistics
Are these two isolated cases representative of the overall statistics regarding how these two groups of people are generally treated by the criminal justice system, or are they just two examples deliberately picked to create a false picture ?
(two) Numerous Variables
No two crimes are exactly the same; no two offenders are exactly the same. Therefore, no two sentences for the 'same' crime can necessarily expected to be equal.
If two separate people, in two separate cases, are charged with, what appears on paper to be the 'same' crime ,the events involved in the enactment of one of these crimes may be significantly different from the events involved of the enactment of the other. They may be of a greater or lesser degree of 'seriousness' in comparison to each other (for instance, one case of 'assault' might have just involved slapping somebody slapping somebody hard on the face once, leaving a bruise, whereas another case of 'assault' might have involved repeatedly hitting somebody with a hammer, breaking several bones). There may be mitigating circumstances in one case, that there isn't in the other. All these factors will have an effect on what kind of sentence somebody will receive..
There are also numerous variables regarding the offender themselves. Had their behaviour in that instance been completely 'out of character' ? Had they been suffering from mental health problems or depression ? Have they got no previous convictions, or are they a regular repeat offender ? Have they shown remorse for their crime ? All these factors will have an effect on what kind of sentence somebody will receive.
Even if, in two separate cases of the 'same' crime, all these numerous variables were exactly the same (which is highly unlikely) that still wouldn't guarantee that two offenders would get exactly the same sentence. The law does not work on the basis of a precise algorithm that has predicted in advance all these potential variables & can therefore 'compute' an exact sentence on that basis. An individual judge just has to consider all these variables, look at sentencing guidelines, & use his / her experience to determine what they consider to be an 'appropriate' sentence accordingly.
Even in that hypothetical situation of all variables being absolutely equal in two separate cases, one cannot reasonably expect two different judges to automatically deliver the exact same sentence. As I said, the law is not an algorithm (and it would be disastrous if it was).
Given all the above factors, it would therefore be illogical, for example, to look at two stories in this week's news, and say "A muslim was charged with assault, and he only got 2 weeks in jail, but that white guy who got charged with assault gets a two year sentence, therefore it's clear that the legal system favours muslims above white people, and muslims always get lesser sentences.". Without understanding, and taking account of, all the variables that would have led to these two different sentences, making an assumption like that is nothing more than bigotry, conspiracy theorising, & confirmation bias.