Discrimination biases

Unwanted arbitrariness is the most important threat for a rational ethic. An important category of unwanted arbitrariness, is discrimination, with the three most known and important examples: sexism, racism and speciesism. Discrimination is fundamentally immoral and should be avoided. However, even people fighting against sexism, racism and speciesism are sometimes vulnerable to discrimination biases. As a result of these biases, they risk discrimination themselves. In this article I present three discrimination biases and give examples of these biases for sexism, racism and speciesism. The goal of this article is to improve rational, critical thinking that helps in the fight against all kinds of discrimination.

Defining discrimination

First we have to define discrimination. Here I define it with three conditions: 1) privileging A over B, 2) based on arbitrary criteria and 3) without tolerating swapping positions. Privileging A over B means treating A better than B in a way that B cannot consistently want (if B were rational and well-informed). Arbitrariness of criteria means that no rule was followed to select the criterion. Swapping positions means treating A like B and vice versa.

A prime example of an arbitrary criterion, is group membership. Consider the group of everyone. This group can be divided in several subgroups, which can be further divided in subsubgroups and so on. Sexists, racists or speciesists select respectively the subgroups of males, whites and humans that are defined by biological characteristics such as sex (in particular genitals), race (in particular skin-color) and species. They could have selected other subgroups (such as females, blacks and pigs), and they could have selected other biological characteristics (such as voice pitch, hair color and biological phylum). As no rules were followed to selected the specific subgroup, or to select a subgroup instead of the group or a subsubgroup, or to select the kind of biological characteristic, the selections made by sexists, racists and speciesists are arbitrary.

When A and B in the definition of discrimination refer to subgroups, a privilege can be detected in terms of a statistical difference between the A and B groups. Statistical discrimination is present when the actual shares (e.g. shares of the total number of crime victims, shares of the total income, shares of the positions of power) differ from the population shares (the relative number of individuals in the subgroups) and this difference cannot be explained by morally relevant criteria (such as personal preferences, degree of innocence,…). Statistical discrimination can be detected by considering a counterfactual ‘blind’ society (e.g. gender blind, color blind, species blind) where the biological characteristic was erased or not visible, or a uniform society, where everyone belonged to the same sex, race and species and is then randomly assigned a sex, race or species. If the pattern in such a blind or uniform society would be different from the actually observed pattern, and if this pattern difference cannot be explained by relevant factors, then statistical discrimination is present. Statistical discrimination can most easily be explained by some people having discriminatory attitudes and behavior. Hence, if statistical discrimination is present, then this is evidence that at least some people discriminate.

To analyze the discrimination biases, we have to consider two distinctions: first the distinction between the undeserved privileged and the undeserved disprivileged (where ‘undeserved’ means not based on morally relevant factors), and second the distinction between two subgroups: the ingroup or A-group and the outgroup or B-group. With respect to sexism, racism and speciesism, the A-group are respectively the men, whites and humans, the B-group are the women, non-whites (e.g. blacks) and non-humans (e.g. pigs). A person from the A-group can be either privileged or disprivileged (I will omit ‘undeserved’ for simplicity), and the same goes for someone from the B-group, so there are four possibilities (privileged A, disprivileged A, privileged B and disprivileged B). There is no statistical discrimination if the shares of privileged are the same in both the A-group and the B-group.

Discrimination neglect

The first discrimination bias is the simplest to understand, as it simply involves not seeing or acknowledging some kind of discrimination. Anti-discrimination activists notice one kind of statistical discrimination, namely the discrimination of disprivileged B (e.g. women, black people), but not the discrimination of disprivileged A (e.g. men, white people). Discrimination neglect is similar to confirmation bias (accepting information that confirms one’s prior beliefs and rejecting contradicting evidence): when the data indicate statistical discrimination of the usually considered disprivileged group B, the data are accepted. When the data point towards discrimination of group A, they are rejected or neglected.

Discrimination neglect of a statistical discrimination allows for an easy test: suppose the data were reversed (i.e. the shares of subgroups A and B were interchanged). If in that case the anti-discrimination activists would put the issue on their agenda, there is statistical discrimination neglected by the anti-discrimination activists.

The most examples of discrimination neglect are examples of statistical sexism. Anti-sexists such as feminists focus on examples where women are disadvantaged relative to men and often strongly neglect examples where men are disadvantaged relative to women. There are several such examples of reverse sexism, where men are discriminated (e.g. child custody, paid parental leave, retirement age, military draft, rescue operations, unemployment rates).

Concerning statistical racism, clear examples are more difficult to find. One example could be violent crimes: US violent crime statistics (2018) show that white on black violent crimes are ten times less likely than expected based on population shares, whereas black on white crimes are 1,4 times more likely than expected. Whites are 9% more likely to be victim of a violent crime, whereas blacks are 17% less likely, compared to a color-blind society. Hence, there is more violence with black perpetrators and white victims than vice versa, contradicting the white supremacy view. This data could indicate statistical discrimination, at least if white and black victims are equally innocent and white and black perpetrators are equally guilty (for example if whites are oppressing blacks, blacks could retaliate by more violent crimes, but whites are more guilty of racism).

I am not aware of examples of discrimination neglect when it comes to speciesism.

The problem with discrimination neglect, is of course that it can perpetuate certain kinds of discrimination.

Privileged-disprivileged bias

The privileged-disprivileged bias can be considered as a specific kind of discrimination neglect that can occur in situations where the A-group and B-group have the same mean value of a property (e.g. wealth, status), but the A-group has a larger spread (i.e. a larger variance or wider statistical distribution) than the B-group. The property can take values from high to low and hence constitutes a hierarchy. The larger spread means that we see more A’s than B’s at the top of the hierarchy, but also more A’s than B’s at the bottom. The privileged-disprivileged bias is a focus on either the top or the bottom (usually the top). A is judged to be privileged, because there are more A’s than B’s at the top. The disprivileged A’s at the bottom are neglected. Similarly, when A is judged to be disprivileged due to the larger share of A’s at the bottom, the privileged A’s at the top are neglected.

There are several examples of privileged-disprivileged bias when it comes to sexism, because it happens that men have a wider statistical distribution for many properties. Consider wealth, and in particular the possession of real estate. Most of the people at the top, i.e. people living in large villa’s or castles and having secondary residences, are men. But also most people at the bottom, i.e. the homeless, are men. The same goes for job status. Most people with the highest status jobs (e.g. ministers and managers), are men, but also most people with the lowest status jobs (the dirty and dangerous jobs such as garbage collector or foot soldier), are men. Other examples: the people with the most sexual partners, are men, the people with the fewest sexual partners, are men. Most perpetrators of violent crimes, are men, most victims of violence, are men. Most people with ecstatic, luxurious lives, are men, most people who commit suicide, are men. The larger variance of men can perhaps be explained by their higher risk taking behavior or the larger variance in cognitive abilities such as IQ.

A related privileged-disprivileged bias is observed in speciesism. Individuals that cause the most animal suffering, are mostly humans, individuals that are most helpful to animals, are also mostly humans. Compared to humans, non-human animals are not so good at helping other animals. Humans can be both animal’s worst enemies and animal’s best friends.

The problem with privileged-disprivileged bias is that it is a hasty generalization that could be harmful to the people in the worst-off positions, the most disprivileged people, by stereotyping them as being privileged. This can happen when speaking about ‘male privilege’, which risks trivializing or disacknowledging the disadvantages experienced by those men who are worst-off. One could say men are privileged because most people at the most privileged positions are men, but one could equally say that men are disprivileged because most people at the most disprivileged positions are men. Anti-sexist activists who focus on male privilege might become too misandrous (men hating). Similarly, anti-speciesist animal rights activists who focus on the harms caused by humans, might become too misanthrope.

Perpetrator-focused discrimination bias

Discrimination neglect (and privileged-disprivileged bias) deals with accepting or neglecting empirical data and is therefore a judgment bias about empirical facts. In contrast, perpetrator-focused discrimination bias is a judgment bias about normative values. The former deals with factual beliefs, the latter deals with moral beliefs.

The perpetrator-focused discrimination bias is the judgment that harm caused by someone from the A-group is worse than harm caused by someone from the B-group. When the perpetrators are A’s, their harm is considered worse than if they were B’s. This results in a kind of discrimination, where the B-group receives a kind of privilege, i.e. a permission to cause more harm, or a sentence reduction (a smaller punishment). A moral rule like “harm caused by A is morally worse than harm caused by B”, explicitly refers to arbitrary groups A and B, and this discriminates A against B. To avoid discrimination, moral rules should never explicitly refer to arbitrary groups. We can also see that such a perpetrator focus and such moral rules involve discrimination, by listening to the victims. Assuming those victims do not have discriminatory attitudes themselves, for them it does not matter whether their harm was caused by an A or a B. If the victims do not make a value distinction and do not have a preference for the group membership of their perpetrators, then no-one should make such a value distinction. We have to care about what the victims care about, and the victims do not care about the sex, race or species of their perpetrators.

An example of this bias in sexism, is criminal sentencing. Men receive 60% higher sentences than women for equal crimes. Arrested women are more likely to avoid convictions and are twice as likely to avoid incarceration if convicted. This is confirmed by other studies. The latter research by Sigrid von Wingerden in the Netherlands indicates that when a woman kills a man, the sentence is 1,6 years lower than when a man kills a man. It could hypothetically be that men are more responsive to punishments than women, which means that higher sentences for men is an effective policy to reduce crimes. But I am not aware of strong evidence in favor of this hypothesis. Hence, this difference in criminal sentencing is likely to involve sexism. If a man is the perpetrator, it is apparently worse than when a woman is the perpetrator. Female perpetrators become more privileged with such a sexist sentencing policy.

An example concerning racism, is the difference in media attention about killings. Blacks killed by whites (think about fatal police shootings in the US) receive more media attention than blacks killed by blacks, although blacks are almost 7 times more likely to be killed by blacks than by whites. Even more extreme: if society were color blind (as if there were only one race), and the number of blacks killed by blacks remained the same, the number of blacks killed by whites would be 30 times lower than the number of blacks killed by blacks. In a color blind media, one would therefore expect 30 times more reporting of killings by black perpetrators than by white perpetrators. This discrepancy in media attention is an indication that the death of a black person is considered worse when the perpetrator is white person (e.g. a white cop) than when the perpetrator is black. Assuming the victim is not racist, for the victim the skin color of the perpetrator is irrelevant.

The most painfully ironic example of perpetrator-focused discrimination bias relates to speciesism. Many anti-speciesist animal rights activists are predation supporters, which means they believe that predation in nature without human interference is never morally bad and that human interventions in nature that knowingly eliminate predation can never be good (not even when these interventions would increase aggregate welfare of wild animals). Those animal activists are against hunting by a human hunter, but condone or support hunting by a lion, even if that lion never respects the most basic animal welfare laws, causes more panic to prey animals and kills more prey animals than a human hunter. The animal activists often explicitly claim that human-caused animal suffering is worse than animal-caused animal suffering. Hence, they explicitly refer to an arbitrary species in their moral rule, and this leads to speciesism. The suffering wild animals don’t care about who causes the suffering. They simply don’t want this extreme unnecessary suffering, and for them it doesn’t matter if it is caused by humans or by non-humans. The predation supporting animal activist is even against research on how to safely and effectively intervene in predation such that wild animal welfare increases, claiming that such interventions violate the autonomy of predators and consist of anthropocentric speciesist human arrogance. This is most ironic, because by catching prey animals, the predator takes away all of the autonomy of those prey animals, and neither the predator nor the prey have this speciesist belief that species membership of the perpetrator is morally relevant. When arrogance means imposing one’s own beliefs or values on others, by imposing their own speciesist moral rule on prey animals, the predation supporting animal activists themselves are being arrogant.

A mirror image of perpetrator-focused discrimination bias is victim-focused discrimination bias, whereby the harm done to a victim of the B-group is considered worse than the harm done to a victim of the A-group.

Three examples of this bias relate to sexism. First, genital mutilation, whereby non-therapeutic, unanesthetized neonatal genital mutilation of girls is strongly prohibited and rejected but the mutilation of boys is mostly condoned. Second, shelters for victims of domestic violence, whereby the ratio of number of shelters for female victims to shelters for male victims is orders of magnitude larger than the ratio of female victims of domestic violence to male victims of domestic violence. Third, criminal sentencing, whereby a man who kills a woman gets a longer prison sentence than a man who kills a man. Concerning racism, we can note that the ratio of the number of media reports about a black person killed by a police officer to the number of media reports about white victims of police violence is larger than the ratio of the number of black people killed by cops to the number of white people killed. Hence, if the victim is black, that victim receives relatively more media attention.

These examples suggest that when a man or a white person is harmed (killed, genitally mutilated,…) it is apparently not as bad as when a woman or a black person is harmed. However, this does not yet imply real discrimination.

First, as with affirmative action, one can argue that this differential treatment of victims serves to rectify another discrimination by giving more privileges to members of the disadvantaged group. If women and black people are generally disadvantaged, measures to target perpetrators of female and black victims can be justified, because those measures advance the positions of the worst-off, namely the disadvantaged women and blacks. Note that with perpetrator-focused discrimination bias, women or black people also gained a privilege, but it concerned an advantage to harm others, and such a privilege is not justified to correct for other disprivileges.

Second, with perpetrator-focused discrimination, the victim did not share the discriminatory attitude. For the victim there is no difference between being a victim of an A-group perpetrator or a B-group perpetrator. But for victim-focused discrimination bias, we have to look at the attitudes of the perpetrator instead of the victim. The perpetrator can be a real sexist, racist or speciesist. The victim-focused discrimination (e.g. stronger punishments for A-group perpetrators who harm B-group victims) can serve to counteract this discriminatory attitude of the perpetrator.

Note: these two considerations also imply that some of the examples of reverse sexism that I gave here, are not necessarily really sexist. Arguments in favor of some differential treatments of men and women could be given, such that the related statistical differences do not yet indicate statistical discrimination.

Advertentie