Group analysis and logical fallacies

[bigfield]

As Togo has pointed out in another thread:

That's also why there's a political split here. Right-wingers tend to emphasise personal responsibility a lot more than the left, and thus tend to see descrimination as person A harming person B. Thus any action taken must be absolutely equal at the individual level, and focusing on helping groups that suffer the most is inherently unequal, and thus morally wrong. Left-wingers tend to focus on group responsibility, and thus tend to see descrimination as group A harming group B. Thus any action that produces an unequal set of outcomes is inherenetly unequal, and thus morally wrong.

There are some fallacies that are typical in arguments that focus on groups as described above:

Ecological fallacy: One compares Group A to Group B, and finds that Group A scores higher on some metric. One infers from that comparison that any given individual from A is more likely to score higher on that metric than any given individual from B. This fallacy is frequently committed in discussions about privilege, whereby exponents of the theory make overly broad and unsupported assumptions about any given individual based on one or more groups they belong to.

Illicit Minor: Identifying that all members of Group A are members of Group B, and that all members of B are in C, and then concluding that all members of C are in Group A. For example, one identifies that a particular parliament was responsible for enacting a law that discriminates between men and women, and then identifies that all members of that parliament were men, and then concludes that all men are responsible for the law. Typically, this fallacy appears when a man complains about a law that harms men, and another person responds by pointing out that the law in question was written by men, the implication being that the men harmed by the law, and the man complaining about the law, hold some of the responsibility for the law's creation.

Equivocation fallacy: One describes two different sets of people as if they were the same set of people. For example, using the set of extant white Americans interchangeably with the set of all white Americans, past and present. This fallacy is committed whilst justifying one's desire to hold extant white Americans accountable for the actions of dead white Americans, such as slavery.

Leaping to Conclusions: One compares Group A to Group B, and finds that Group A scores higher on some metric. One attributes this difference to cause X, despite lacking grounds to do so. For example, one identifies the gender pay gap and attributes it to sexism without further analysis.

Obviously, a political position based on these fallacies is unsound. If a person is going to claim that Group A is harming Group B, when one needs to provide a sound argument for their choice of categories, free of fallacies like the ones listed above, or any others for that matter.

ETA: One also needs to show why unequal outcomes are morally wrong -- it should not be an axiom of one's position.

 

[Derec]

Thus any action that produces an unequal set of outcomes is inherenetly unequal, and thus morally wrong.

Thus any method to select players for the NBA that doesn't result in 64% of players to be non-hispanic white and only 13% to be black is "inherenetly[sic] unequal and thus morally wrong".

 

[dismal]

Thus any action that produces an unequal set of outcomes is inherenetly unequal, and thus morally wrong.

Thus any method to select players for the NBA that doesn't result in 64% of players to be non-hispanic white and only 13% to be black is "inherenetly[sic] unequal and thus morally wrong".

Since the NBA draws from the global population there should be more Chinese and Indians. Also women, old people and children seem underrepresented.

 

[Togo]

Obviously, a political position based on these fallacies is unsound. If a person is going to claim that Group A is harming Group B, when one needs to provide a sound argument for their choice of categories, free of fallacies like the ones listed above, or any others for that matter.

ETA: One also needs to show why unequal outcomes are morally wrong -- it should not be an axiom of one's position.

I'm not sure these are true. A logical argument is one that safely takes you from premise A to conclusion B through a series of logically sound steps. But political and moral positions are more often a difference in values (i.e. premises) rather than differences in logic. Avoiding axioms is literally impossible.

For example, you may feel that treating people equally is important. And you may feel that helping the poor is important. Giving bread to starving people fulfils the second value by violating the first. Giving free lambourginis to starving people also fulfils the second value by violating the first. The question then becomes how you balance and reconcile competing values.

I'd love to think that everyone who disagrees with me is simply illogical, but I very doubt it's that simple.

 

[arkirk]

Obviously, a political position based on these fallacies is unsound. If a person is going to claim that Group A is harming Group B, when one needs to provide a sound argument for their choice of categories, free of fallacies like the ones listed above, or any others for that matter.

ETA: One also needs to show why unequal outcomes are morally wrong -- it should not be an axiom of one's position.

I'm not sure these are true. A logical argument is one that safely takes you from premise A to conclusion B through a series of logically sound steps. But political and moral positions are more often a difference in values (i.e. premises) rather than differences in logic. Avoiding axioms is literally impossible.

For example, you may feel that treating people equally is important. And you may feel that helping the poor is important. Giving bread to starving people fulfils the second value by violating the first. Giving free lambourginis to starving people also fulfils the second value by violating the first. The question then becomes how you balance and reconcile competing values.

I'd love to think that everyone who disagrees with me is simply illogical, but I very doubt it's that simple.

The problem with this type of logical examination of human behavior is simply that human behavior is NEVER ALL THAT IS HAPPENING. In political discussions we tend to withdraw into a simplified vacuum in which there are only humans and their posessions. Actually our political realities are the result of a greater field of action than any of our groups seem capable of completely understanding. Instead of labeling political factions Groups A and B and C, we could label them Misunderstanding A and B and C. In all fairness to the human race, it is not misunderstanding so much as short understanding. So what am I talking about...perhaps it is the human capacity when faced with intractable problems to become frustrated with the others and sort them out into Groups B and C. Our political ideologies must NOT PAY HOMAGE to our capacity to become frustrated and write off others with very great ideological differences with us.

When you see someone who is hungry and you have plenty of bread, you actually do not go through any principles analysis whether you give him some bread or not. It is a matter of how much empathy you have for that person. A lot of the "analyses" in this thread are burdened with logical structures that do not take into account how our brains function. That was a big part of the problem with Aristotelian Logic. That is a big part of the problem with any grammar based system of reasoning. We may be attempting to communicate and making a great effort to do so and still be failing to clearly communicate.

That is why I feel the best approach is to adopt a more humble and humanistic attitude toward people one does not understand and to always weigh their experiences (which we cannot completely know) and their leaders' ability to lie and to be frustrated, and in fact manifest all the same negative qualities we can be capable of manifesting ourselves.

 

[ronburgundy]

IOW, understanding people or defining their traits based upon on what racial or gender group they belong to is racist and sexist.

 

[fromderinside]

IOW, understanding people or defining their traits based upon on what racial or gender group they belong to is racist and sexist.

So how does one determine racial or gender group? One needs to know that to make the evaluation you spouted doesn't one? ...and if ones spouted group specification is invalid, then what? Isn't the cat out of the bag? Ah. Politics.

 

[dismal]

IOW, understanding people or defining their traits based upon on what racial or gender group they belong to is racist and sexist.

So how does one determine racial or gender group? One needs to know that to make the evaluation you spouted doesn't one? ...and if ones spouted group specification is invalid, then what? Isn't the cat out of the bag? Ah. Politics.

Careful, if there aren't racial and gender groups there can't be racial and gender grievances.

Then what are we going to do here?

We may have to shut the forum down.

 

[Jolly_Penguin]

This Op is excellent and should be a permament sticky for the forum. I have seen examples here of each of these logic fails. Being a "free thought' forum, we should keep logic and reason in the forefront, especially when discussions get heated, as they often do.

 

[bigfield]

Obviously, a political position based on these fallacies is unsound. If a person is going to claim that Group A is harming Group B, when one needs to provide a sound argument for their choice of categories, free of fallacies like the ones listed above, or any others for that matter.

ETA: One also needs to show why unequal outcomes are morally wrong -- it should not be an axiom of one's position.

I'm not sure these are true. A logical argument is one that safely takes you from premise A to conclusion B through a series of logically sound steps. But political and moral positions are more often a difference in values (i.e. premises) rather than differences in logic. Avoiding axioms is literally impossible.

My point was not that axioms are unavoidable, but specifically that there is good reason to avoid that axiom in particular.

Derec used the example of the NBA above, where some racial groups are under-represented compared to their representation in the US or world populations. There are non-discriminatory reasons why most NBA players are black, and therefore one should not presuppose that this racial makeup is a morally-wrong outcome.

For example, you may feel that treating people equally is important. And you may feel that helping the poor is important. Giving bread to starving people fulfils the second value by violating the first. Giving free lambourginis to starving people also fulfils the second value by violating the first. The question then becomes how you balance and reconcile competing values.

The example you've provided demonstrates that in order for a value system to be practicable, it needs to have more nuance than the above.

Rather than simply stated that all people should be treated equally, one could state that all people should be treated according to the same algorithm (or decision-making process). The algorithm should only use a person's relevant attributes when deciding how to treat them.

For example, race is irrelevant when determining which job candidate to hire.

With respect your example, this means that the same algorithm should be applied to every individual when determining whether to give them money or take money from them. If an individual is below a minimum income threshold, then apply a negative tax rate, and if the person is above a tax-free threshold, then apply a positive tax rate. All individuals are subject to the same rules, and if a person's income changes, then their treatment change accordingly.

What makes a system fair is that like cases are treated alike, and different cases are treated differently, but differentiating cases based on irrelevant attributes is not fair and not morally right.

 

[bigfield]

I'm not sure these are true. A logical argument is one that safely takes you from premise A to conclusion B through a series of logically sound steps. But political and moral positions are more often a difference in values (i.e. premises) rather than differences in logic. Avoiding axioms is literally impossible.

For example, you may feel that treating people equally is important. And you may feel that helping the poor is important. Giving bread to starving people fulfils the second value by violating the first. Giving free lambourginis to starving people also fulfils the second value by violating the first. The question then becomes how you balance and reconcile competing values.

I'd love to think that everyone who disagrees with me is simply illogical, but I very doubt it's that simple.

The problem with this type of logical examination of human behavior is simply that human behavior is NEVER ALL THAT IS HAPPENING. In political discussions we tend to withdraw into a simplified vacuum in which there are only humans and their posessions. Actually our political realities are the result of a greater field of action than any of our groups seem capable of completely understanding. Instead of labeling political factions Groups A and B and C, we could label them Misunderstanding A and B and C. In all fairness to the human race, it is not misunderstanding so much as short understanding. So what am I talking about...perhaps it is the human capacity when faced with intractable problems to become frustrated with the others and sort them out into Groups B and C. Our political ideologies must NOT PAY HOMAGE to our capacity to become frustrated and write off others with very great ideological differences with us.

When you see someone who is hungry and you have plenty of bread, you actually do not go through any principles analysis whether you give him some bread or not. It is a matter of how much empathy you have for that person. A lot of the "analyses" in this thread are burdened with logical structures that do not take into account how our brains function. That was a big part of the problem with Aristotelian Logic. That is a big part of the problem with any grammar based system of reasoning. We may be attempting to communicate and making a great effort to do so and still be failing to clearly communicate.

That is why I feel the best approach is to adopt a more humble and humanistic attitude toward people one does not understand and to always weigh their experiences (which we cannot completely know) and their leaders' ability to lie and to be frustrated, and in fact manifest all the same negative qualities we can be capable of manifesting ourselves.

A logical approach is extremely useful for determining political policy and system design. Human decision-making is rife with cognitive and is not a good model for running an organisation, whether it be a government or private institution.

For instance, the best way to select a job candidate may be to delegate the job to a computer, thus making it impossible for a human evaluator to make a sub-optimal decision based on either conscious or unconscious biases.

 

[Blahface]

One group fallacy that I hate is when conflating the beliefs and actions of multiple subgroups onto the super-group and act like the super-group has all these subgroup properties.

An example of this is somebody criticizing atheists for generalizing and making blanket statements about Christians while they are also criticizing people who make blanket statements about Muslims as if atheists have a double standard. I think it is more likely that the atheists who are criticizing Christians are the same atheists who are criticizing Muslims. When a different set of atheists say that we shouldn't generalize about Muslims, then they get blamed for applying a double standard because they are all seen as apart the atheist super-group.

Another example of this is when Sam Harris talks about polls in Muslim countries. He cites the worst of each country and leaves the impression of a Muslim super-villain.

 

[tupac chopra]

Rather than simply stated that all people should be treated equally, one could state that all people should be treated according to the same algorithm (or decision-making process). The algorithm should only use a person's relevant attributes when deciding how to treat them.

For example, race is irrelevant when determining which job candidate to hire.
.

Except that you haven't mentioned any relevant attributes only one irrelevant one

 

[Togo]

For example, you may feel that treating people equally is important. And you may feel that helping the poor is important. Giving bread to starving people fulfils the second value by violating the first. Giving free lambourginis to starving people also fulfils the second value by violating the first. The question then becomes how you balance and reconcile competing values.

The example you've provided demonstrates that in order for a value system to be practicable, it needs to have more nuance than the above.

Rather than simply stated that all people should be treated equally, one could state that all people should be treated according to the same algorithm (or decision-making process). The algorithm should only use a person's relevant attributes when deciding how to treat them.

I'm not sure this 'single process' really adds anything.

For example, in process A, you use a different hiring process depending on whether you're hiring a black man or a white one. In process B you use a single hiring process that forks depending on whether the person is black or white. Is there really a difference here?

More relevant perhaps is the insistence on ignoring certain attributes entirely. But that approach has it's own problems. To take an extreme case, a motor car company had problem with lateness it's Mexican factory. Some employees would turn up on time, and some were persistently late. The situation got quite ugly. People who pointed out the tardy employees were Mexican, while those arriving on time were German, were accused of discrimination. It took some external consultants being called in to discover that the German employees were housed in a hotel next to the factory, while Mexican employees were walking in from several miles away. The local road had been washed away, and they were literally fording rivers before dawn each morning in an effort to get into work.

It's all very well saying you should be purposefully blind to certain attributes, but it produces some very perverse results in practice.

With respect your example, this means that the same algorithm should be applied to every individual when determining whether to give them money or take money from them.

Which was precisely the reason given for why the company couldn't pay to fix the local road to the factory. Because many employees didn't use that road, and thus would be unfairly charged for a road they didn't need

What makes a system fair is that like cases are treated alike, and different cases are treated differently, but differentiating cases based on irrelevant attributes is not fair and not morally right.

And the attributes that we consider, and how we consider them, comes straight back to competing values again.

For example, let's say you've worked out certain traits in your employees that mean they tend, as a group, to have lower absenteeism, less chance of employee theft, and good relationships with other workers. Is it fair to rate that trait highly, and favour hiring people who have it? The answer doesn't depend on what effect that trait has on your business, but whether the trait is something like education level, or whether it's something more like membership of the freemasons. That's not a logical or business judgement, but a values judgement about what sort of traits ought to matter.

 

[ronburgundy]

IOW, understanding people or defining their traits based upon on what racial or gender group they belong to is racist and sexist.

So how does one determine racial or gender group? One needs to know that to make the evaluation you spouted doesn't one? ...and if ones spouted group specification is invalid, then what? Isn't the cat out of the bag? Ah. Politics.

By using evidence about each individual directly tied to the very specific traits that define those categories.
IOW, individual-level evidence defines what category they belong to on that particular variable. That is not problematic because it does not infer/presume anything about any person based upon group membership. Rather it merely defines membership based upon evidence about each specific person. It is only when you take other features correlated with but not defining aspects of the categories and presume that these aggregated relations/differences apply at the individual member level that you run into problems and engage in what is the very definition of psychological racism and sexism.

 

[arkirk]

The problem with this type of logical examination of human behavior is simply that human behavior is NEVER ALL THAT IS HAPPENING. In political discussions we tend to withdraw into a simplified vacuum in which there are only humans and their posessions. Actually our political realities are the result of a greater field of action than any of our groups seem capable of completely understanding. Instead of labeling political factions Groups A and B and C, we could label them Misunderstanding A and B and C. In all fairness to the human race, it is not misunderstanding so much as short understanding. So what am I talking about...perhaps it is the human capacity when faced with intractable problems to become frustrated with the others and sort them out into Groups B and C. Our political ideologies must NOT PAY HOMAGE to our capacity to become frustrated and write off others with very great ideological differences with us.

When you see someone who is hungry and you have plenty of bread, you actually do not go through any principles analysis whether you give him some bread or not. It is a matter of how much empathy you have for that person. A lot of the "analyses" in this thread are burdened with logical structures that do not take into account how our brains function. That was a big part of the problem with Aristotelian Logic. That is a big part of the problem with any grammar based system of reasoning. We may be attempting to communicate and making a great effort to do so and still be failing to clearly communicate.

That is why I feel the best approach is to adopt a more humble and humanistic attitude toward people one does not understand and to always weigh their experiences (which we cannot completely know) and their leaders' ability to lie and to be frustrated, and in fact manifest all the same negative qualities we can be capable of manifesting ourselves.

A logical approach is extremely useful for determining political policy and system design. Human decision-making is rife with cognitive and is not a good model for running an organisation, whether it be a government or private institution.

For instance, the best way to select a job candidate may be to delegate the job to a computer, thus making it impossible for a human evaluator to make a sub-optimal decision based on either conscious or unconscious biases.

You mean the boss just delegates the sub-optimal decision to the computer? Computers do what we tell them to. The issue does not vanish just because a machine makes decisions. The problem is the failure to provide some form of accommodation for all people so that all people are afforded the opportunity to contribute. To limit human employment is sub-optimal in itself.

There can be my logic and your logic and we can declare each others' logical process illogical. Your premise in the assignment of hiring decisions to computers completely overlooks the fact that: (1) One may program his prejudices into the decision making functions of his computer. and (2) All complex devices yield results which ultimately require human oversight for appropriateness.

The error in your system may be in the limitations of your private operation...its failure to provide sufficient engagement (employment) of the society in which it operates. The internal computations of your company computer says everything is fine inside the white picket fence your programming has erected around your organization, within your realm of control. It is unlikely you will program your computer to analyze the opportunity costs to society your organization's policies shift onto society at large.

 

[bigfield]

The example you've provided demonstrates that in order for a value system to be practicable, it needs to have more nuance than the above.

Rather than simply stated that all people should be treated equally, one could state that all people should be treated according to the same algorithm (or decision-making process). The algorithm should only use a person's relevant attributes when deciding how to treat them.

I'm not sure this 'single process' really adds anything.

For example, in process A, you use a different hiring process depending on whether you're hiring a black man or a white one. In process B you use a single hiring process that forks depending on whether the person is black or white. Is there really a difference here?

My point is that there should be no fork depending on whether the person is black or white.

More relevant perhaps is the insistence on ignoring certain attributes entirely. But that approach has it's own problems. To take an extreme case, a motor car company had problem with lateness it's Mexican factory. Some employees would turn up on time, and some were persistently late. The situation got quite ugly. People who pointed out the tardy employees were Mexican, while those arriving on time were German, were accused of discrimination. It took some external consultants being called in to discover that the German employees were housed in a hotel next to the factory, while Mexican employees were walking in from several miles away. The local road had been washed away, and they were literally fording rivers before dawn each morning in an effort to get into work.

It's all very well saying you should be purposefully blind to certain attributes, but it produces some very perverse results in practice.

With respect your example, this means that the same algorithm should be applied to every individual when determining whether to give them money or take money from them.

Which was precisely the reason given for why the company couldn't pay to fix the local road to the factory. Because many employees didn't use that road, and thus would be unfairly charged for a road they didn't need

The race of the workers is irrelevant to this example.


It is entirely fair for a company to have a policy where they invest money in an employee to keep them at the company despite exceptional circumstances. This is the same principle as is behind parental leave and extended sick leave.

The fact that some employees in your example were Mexican, and some were German, is not relevant.

What makes a system fair is that like cases are treated alike, and different cases are treated differently, but differentiating cases based on irrelevant attributes is not fair and not morally right.

And the attributes that we consider, and how we consider them, comes straight back to competing values again.

For example, let's say you've worked out certain traits in your employees that mean they tend, as a group, to have lower absenteeism, less chance of employee theft, and good relationships with other workers. Is it fair to rate that trait highly, and favour hiring people who have it?

Assuming that one is not committing an ecological fallacy and that these hypothetical group averages do actually reflect the likelihood of traits in an individual, then yes, it belongs in the evaluation.

The answer doesn't depend on what effect that trait has on your business, but whether the trait is something like education level, or whether it's something more like membership of the freemasons.

Of course it depends on the effect on the business: if it didn't then it would be relevant to the evaluation.

That's not a logical or business judgement, but a values judgement about what sort of traits ought to matter.

If the traits are not relevant to the candidate's ability to do the job, then they should not affect the evaluation.

 

[arkirk]

Bigfield: Do you have an appropriate algorithm for morality? I noticed you mentioned it in your post immediately above this one. In my estimation having read up on the so called ecological fallacy and studied a few examples, it really should be called the statistical fallacy.
You still have not told me how you would be able to avoid the "cognitives" as you call them by merely programming them into your human resources computer.

 

[bigfield]

A logical approach is extremely useful for determining political policy and system design. Human decision-making is rife with cognitive and is not a good model for running an organisation, whether it be a government or private institution.
For instance, the best way to select a job candidate may be to delegate the job to a computer, thus making it impossible for a human evaluator to make a sub-optimal decision based on either conscious or unconscious biases.

You mean the boss just delegates the sub-optimal decision to the computer? Computers do what we tell them to. The issue does not vanish just because a machine makes decisions.
The problem is the failure to provide some form of accommodation for all people so that all people are afforded the opportunity to contribute. To limit human employment is sub-optimal in itself.

I don't see how that is relevant.

There can be my logic and your logic and we can declare each others' logical process illogical. Your premise in the assignment of hiring decisions to computers completely overlooks the fact that: (1) One may program his prejudices into the decision making functions of his computer. and (2) All complex devices yield results which ultimately require human oversight for appropriateness.

Computers are already capable of making some decisions much better than a human is able to do. The problem of designer bias is easily to overcome, just as it is easy to overcome in scientific research.
Human oversight is only required to make sure that the system is functioning as designed; it is not up to the human to judge the appropriateness of each result based on their own subjective biases, because that would render the automated system useless.

The error in your system may be in the limitations of your private operation...its failure to provide sufficient engagement (employment) of the society in which it operates.

What does sufficient engagement entail of society entail?

The internal computations of your company computer says everything is fine inside the white picket fence your programming has erected around your organization, within your realm of control.

The system I described does nothing of the kind.

It is unlikely you will program your computer to analyze the opportunity costs to society your organization's policies shift onto society at large.

Of course it wouldn't analyse that - it is not supposed to. The hypothetical system I described above was a process for evaluating job candidates.

 

[Loren Pechtel]

Rather than simply stated that all people should be treated equally, one could state that all people should be treated according to the same algorithm (or decision-making process). The algorithm should only use a person's relevant attributes when deciding how to treat them.

Exactly. Total equality doesn't work.

For example, race is irrelevant when determining which job candidate to hire.

No. There are times when you definitely care about race:

1) Actors. If you are portraying multiple characters related by blood you want to choose characters that are not obviously not related. Star Wars showed Luke as white, we didn't see Vader at that time. Later when we do see Vader you need him to be approximately white.

2) Actors. When dealing with historical things you often have to choose an actor that looks like the character. (Note that "historical" can cover more than history. If you're going to do a Nero Wolfe TV show you need a fat guy to portray Nero Wolfe. While he never actually existed the books are already written, you shouldn't go against them.)

3) Models. Sometimes the color of what is being modeled restricts the color of the model.