Broken Symmetries

Broken symmetries offer powerful insights in social science.  When we treat differently two things that a theory says should be treated similarly, that is often a strong clue that theory is missing something important.  In particular, when a normative theory says we should treat two things similarly, that we actually treat differently, that suggests we do not actually follow this normative theory.

Oddly, many folks feel they are excused for treating apparently similar things differently if they can point to any feature distinguishing those things:

  1. “You you do X in case A but do Y in case B, yet A and B do not differ in the features F,G,H that your theory T says are relevant for cases A and B.”
  2. “Ah, but cases A and B differ on feature R.”
  3. “But your feature R is not obviously relevant in your theory T.”

Except person 2 rarely anticipates critique 3; they seem to think it sufficient if it is logically possible that feature R may somehow be connected to the theory’s features F,G,H.

Here is an example from Bennett Haselton reviewing Steven Landsburg’s new book The Big Questions:

Landsburg:  Bert wants to hire an office manager and Ernie wants to manage an office. The law allows Ernie to refuse any job for any reason. If he doesn’t like Albanians, he doesn’t have to work for one. Bert is held to a higher standard: If he lets it be known that no Albanians need apply, he’d better have a damned good lawyer.

These asymmetries grate against the most fundamental requirement of fairness — that people should be treated equally, in the sense that their rights and responsibilities should not change because of irrelevant external circumstances.

Haselton: But I think the laws do treat all people equally, because they apply equally whether Bert is discriminating in deciding whether to hire Ernie, or whether Ernie is discriminating in deciding whether to hire Bert. The laws don’t apply equally to all roles that people play, which is the distinction that Landsburg is highlighting — but laws never apply equally to different roles, since roles are defined by what we do, and what is the point of laws, except to draw distinctions based on behaviors?

It does not seem to occur to Haselton to ask why the different role of employee vs. employer is relevant in our standard theories suggesting we ban ethnic discrimination.  He feels he is done if he identifies any difference between the two cases.

But everything is different somehow; if finding just one difference is enough to excuse treating any two things differently, one will always have excuses for any asymmetric treatment.

GD Star Rating
a WordPress rating system
Tagged as: ,
Trackback URL:
  • Bill

    The relevant distinction is that the employer is transfering property (wages) to the employee, or would be employee, or is engaging in no transaction if he discriminates in refusing to hire for a fair exchange of labor for work. When the employee is deciding not to be hired, he is injuring himself; he is not getting the full value of his labor, but the injury is to him, not to another.

  • http://lyingeyes.blogspot.com ziel

    When the employee is deciding not to be hired, he is injuring himself; he is not getting the full value of his labor, but the injury is to him, not to another.

    You’re assuming employees are powerless actors – mere pawns. But this is certainly not always true. Suppose we have 2 employers in a technical field, one of whom is Jewish, and a talented and coveted prospect who is Jewish, and will only work for the Jewish employer. The gentile employer is clearly being hurt by this discrimination. But the employee is also hurting himself, you might say, by reducing his bargaining power. Quite true.

    But turn the tables around – suppose the gentile owner won’t hire Jews – he’d be hurting himself, too, wouldn’t he? The harm goes in both directions in both situations. You just don’t see it because you’re so conditioned to thinking of discrimination as a one-way harm by all-powerful majority employers against helpless minority job-seekers who have no recourse but to seek employment from these employers. If economic talent were distributed evenly among various groups, discrimination would be a non-issue, because each group could effectively arbitrage the discriminatory environment by “taking care of their own.” Unfortunately, this is not actually the case.

    • magfrump

      You say “economic talent”–I might prefer the term “economic power.”
      That is, employers have power in our status quo, not necessarily due to employers being more “talented” per se as much as due to it being easier to have power as an employer.

      This wouldn’t be true in all possible worlds but since it tends to be in ours it is a worthy consideration for policy (not that it’s actually what motivates policy).

  • http://entitledtoanopinion.wordpress.com TGGP

    I was chomping at the bit to respond to Bill, but ziel did an excellent job of it. Good on ya, and damn you for beating me to it!

    I don’t know if he actually agrees with the normative principle, but Ilkka Kokkarinen gives a pragmatic justification for our behavioral asymmetry.

  • Unnamed

    My take is that Landsburg is saying that it’s odd that there’s this asymmetry between employers and employees, as such asymmetries violate a basic principle of fairness. Haselton is responding that it’s not so odd – there are lots of legal asymmetries between employers and employees, and between landlords and tenants, and between vendors and consumers – laws differ for different roles. That doesn’t explain why this particular asymmetry exists (or why it should exist), but it does suggest that we shouldn’t take this asymmetry to be a violation of the principle of fairness (at least not without further argument).

  • Unnamed

    Here are 3 possible explanations of the asymmetry:

    1. Historically, discrimination by employers was a problem and discrimination by employees was not.
    2. Cases of discrimination by employers can often be identified and cases of discrimination by employees generally cannot. (This is the one in TGGP’s link)
    3. Part of what employees get from a job is the experience of working at that job. This is important to them and it’s subjective. Employers are just paying money to get tasks accomplished. So we should be more hesitant to restrict employees’ employment decisions. (If they say that they don’t like the job, what can we say? Are we going to force them to work there?)

    In support of the #3, note that in many jurisdictions the law forbids employers from firing employees without cause. But I’m not aware of any laws forbidding employees from quitting without cause.

    • http://hanson.gmu.edu Robin Hanson

      Why is wanting a subjective experience rather than tasks accomplished relevant for a theory of why ethnic discrimination is bad?

      • Unnamed

        It’s relevant for whether there should be any laws restricting employees’ choice of employer, including anti-discrimination laws. Employers have various legal restrictions on their choice of employees (including non-discrimination, and, in some cases, only being allowed to fire for cause), and employees have almost none. My suggestion is that part of the reason why we’re more willing to interfere with employers’ choices is that they’re just choosing someone to get tasks accomplished, not something that will fill many hours of their subjective experience each day.

      • JGWeissman

        Isn’t managing employees part of an employer’s subjective experience?

      • Nick Tarleton

        It’s part of a manager’s subjective experience, but most employers don’t have subjective experiences.

      • Constant

        “most employers don’t have subjective experiences.”

        Because they’re reptilian extraterrestrials. But even so, they’re also trying to lower their own costs by giving their employees a better subjective experience, which allows them to offer their employees a lower salary. Because of this, their employees’ subjective concerns are also their concerns, regardless of whether they are soulless space reptoids who only care about money.

        Anyway, this looks like speculative cherry-picking of arguably relevant differences for the purpose of after-the-fact rationalization of a decision long since made, an example of confirmation bias. I could just as easily come up with the reverse. For example, employers care about the bottom line but employees care about subjective experience, and therefore employers are more liable to act in a rational way with respect to the bottom line. Racism is irrational, so employers are unlikely to be racist. Employees, however, care greatly about their subjective experience and so cannot be trusted to behave in a non-racist fashion, and so they must be regulated all the more strongly.

        See how easy that is?

  • George

    It’s hard not to comment on the particular example. To me, these laws are obviously about perceived bargaining power (employers are more like a monopsony, and employees rarely negotiate terms). Surely bargaining power is a relevant external circumstance?

    • Peter Twieg

      Even if employers had this bargaining power, it’s not as though that is a sufficient or necessary condition for their having discriminatory preferences in employees. So it seems like this is just an appeal to another unimportant difference.

      • magfrump

        Say the law is about preventing “injustice.”

        Say that discriminatory practices are unjust.

        Then if a powerful faction (the employer) starts discriminatory action, this will cause injustice.
        If a generally-less-powerful faction starts discriminatory action, this will create generally less injustice.

        In this specific case, the unjust prospective manager will create an opportunity cost of the position not being filled. The position can generally be filled by another manager, which will not cause “injustice,” just delay.

        Also, the main cost will be the inefficiency, which is much less visible, so if laws are actually created to convince people that they are just, they will cure the visible injustice (I don’t want to be fired for being Albanian!) and not the invisible one (Albanians might not get managed for a while).

  • Matt Simpson

    I think Haselton is basically right if we are only considering the principle that the rules must apply to each person equally. No matter the rule, there will always be winners and losers depending on circumstance, but as long as the rule is consistently applied in all cases and doesn’t explicitly single out anyone for special treatment (good or bad), the rule is ‘fair’ in that sense.

    However, this is a very weak requirement that doesn’t fully capture our moral intuitions, which is what Landsburg seemed to be talking about. For example, if a rule specifies the death penalty for petty theft, it’s ‘fair’ in Haselton’s sense, but we don’t think it’s ‘fair’ in the moral sense. Also, the special treatment clause is very slippery. On closer examination, It’s hard not to as forbidding all rules! In the death penalty case, is the rule not singling out those people who commit theft?

  • Mike

    That’s great and all, but the example doesn’t demonstrate the general pattern you give. In the second sentence, he says that “R is a relevant part of the theory”.

  • Pingback: Recomendaciones « intelib

  • Nick

    I teach intro to ethics and my undergrads do this all the time and it drives me crazy.

    • http://hanson.gmu.edu Robin Hanson

      I feel your pain.

  • http://hanson.gmu.edu Robin Hanson

    On enforceability, we often have laws that are more easily enforced in some situations than in others, yet we still leave the law on the books for all the situations, just in case it might be enforceable.

    On negotiating power, both sides tend to have similar amounts, though there are some exceptions.

    • http://cob.jmu.edu/rosserjb Barkley Rosser

      Robin,

      “Some exceptions”? Really?

      Well, partly this depends on the overall state of the economy, with employers having more power in recessions and would-be employees gaining power as the economy does better and the unemployment rate drops.

      However, on average, there is an asymmetry of power. While some would-be employees are employed, many are not, whereas the employer is definitely employed. While there are certain situations where the would-be employee makes more money or has more wealth than the employer (a general manager of a sports team trying to hire a star athlete), much more often than not it is the other way around, as indicated by the already stated point about who is more likely to be employed than whom.

      Of course, Landsburg spends time supposedly demolishing arguments for redistribution, so these sorts of inequalities of income and wealth and economic position (employed versus unemployed) do not mean doodley-squat to him. But they do to most people on ethical grounds.

    • George

      “On negotiating power, both sides tend to have similar amounts…”

      Well, it doesn’t feel like that for most people. My perception might be wrong, but it is only perceptions that can create laws.

      I’ve never heard anyone argue that employees (on average) have as much negotiating power as employers. It seem quite an important thing to understand.

  • Robert Koslover

    Your “symmetry” viewpoint reminds me of some arguments about gay marriage. Example: Both heterosexuals and homosexuals are legally allowed to marry people of the opposite sex. Likewise, both heterosexuals and homosexuals are legally prohibited from marrying people of the same sex. Thus, this is absolutely equal treatment under the law, right? Or… might there be some question and/or debate as to what constitutes “symmetry?”

  • Chibinium

    A circle is symmetrical to a sphere as long as they’re not rotated. The question is whether that rotation is meaningful.

    Another “rotation” in the gay marriage argument is whether marriage is for the purpose of raising children, in which case old couples who haven’t had children would be obligated to return all tax benefits. However, a gay couple adopting would then qualify for marriage, defeating the purpose of the proposal’s adherents!

  • Alex Perrone

    Except person 2 rarely anticipates critique 3; they seem to think it sufficient if it is logically possible that feature R may somehow be connected to the theory’s features F,G,H.

    And of course person 1 doesn’t always remember to respond with critique 3.

  • jonathan

    Without discussing the particular example, the argument being advanced is that we should perhaps assume symmetry and yet we have strong evidence that this is not the way the universe is designed. Physics deals with assymetrical forces and design issues in many areas. Perhaps more importantly for your assumption, nearly the entire history of set theory is based on a lack of symmetry that is built into the axioms. To avoid the technical, without something akin to a theory of types – which directly addresses your example – you have contradiction (which makes illegality, not merely inconvenience). The construction of our mathematics rests in work that requires asymmetry.

    On a practical level, sure one must question whether an asymmetry is sensible but the reason for the asymmetry is absolutely fundamental. It is no accident that we tend to generate asymmetrical results, some rational and others even invidious. If not, you’d be out of a job.

  • Carl Shulman

    Employers are less numerous than employees, so if being forced to transact with people is unpleasant, fewer people will be bothered per transaction enabled by a policy that compels employers than one that compels both employees and employers.

    • http://eucalculia.blogspot.com John Faben

      Is this supposed to be an argument for the status quo? Or an explanation of it? If the latter, I guess that makes sense, but sort of goes against the standard arguments people make it cases like this: employers just aren’t a large enough group to have any political power, so no-one stands up for their rights.

      If the former, you’re just shuffling the asymmetry round. You’re saying that one employer being inconvenience 10 times is just as bad as 1 employee being inconvenienced once.

  • Michael Foody

    I think the question of enforcibility makes an important difference. It is much more difficult to tell whether a prospective employee is discriminating against a potential employer based on inappropriate criteria than it is possible to determine the opposite.
    1) Most employers choose many employees, most employees choose one employer. An employer who is discriminating based on inappropiate criteria might exhibit a pattern that would provide at least circumstantial evidence of discrimination. It is unlikely that an individual would ever exhibit such a pattern.
    2) For employers there is a paper trail documenting the request for employment, for employees there is seldom any documentation proving that the prospective employee was informed about the position.

    If you could prove that a prospective employee discriminated against a prospective employer how would you compensate the prospective employer? Would you force the worker to take a position they prefered not to take? What job a person has has an enormous impact on their life. Who a firm hires has a much smaller impact on anyones life (for the most part). Enforcing antidiscrimination statutes on the part of employees would be much more difficult, and have more potential for abuse, and more negative consequences as far as freedom and utility than enforcing the similar statutes as they apply to employers.

  • http://www.rationalmechanisms.com richard silliker

    When I read this post I was intrigued and decided to look at symmetry in nature and came upon this.

    http://www.livescience.com/strangenews/051221_symmetry_nature.html

  • tgrass

    Accepting symmetry in the employment example does not necessarily require extending protection to the employer. One could as easily conclude there should be no equal opportunity standard for either.

  • Yisong

    The best reason that I can come up with is that the employer is representing the interests of an institution/organization that is somehow more public (since it manages the professional and commercial relationships of multiple people), while the potential employee is representing his or her personal interests only. America has always respected a person’s individual beliefs, so it’s OK when someone’s personal decisions are influenced by things like racism.

  • ben

    Please correct me if I’m missing something, but this doesn’t make any sense to me- Haselton is clearly right, and Landsburg is clearly wrong. Consider, for instance, laws against cell phone use while driving. In either case, the political process has determined that it is in the public interest to proscribe a certain action (ethnic discrimination, talking on a cell phone) under a certain set of circumstances (when hiring people, when driving). If someone doesn’t want their choice of actions restricted, they are perfectly free to avoid putting themselves in that particular circumstance. One can certainly debate whether these kinds of things really are in the public interest, but Landsburg’s implication that any law that applies only in certain circumstances is fundamentally unfair because it doesn’t apply to everyone equally is absurd,

    • Constant

      Yes, ben, you are missing something. You are missing this:

      It does not seem to occur to Haselton to ask why the different role of employee vs. employer is relevant in our standard theories suggesting we ban ethnic discrimination. He feels he is done if he identifies any difference between the two cases.

      In the case of using the cell phone, we indeed should ask why the different activity of driving or standing is relevant. We are not done if we identify just any difference. The difference needs to be relevant.

      • Alex Perrone

        After some thought, I think the problem with the fallacy, which reduces it from a fallacy to a pattern of commonly poor argumentation, is with critique 3:

        “3. But your feature R is not obviously relevant in your theory T.”

        should be changed to

        “3. But your feature R is not obviously relevant in *my* theory T.”

        Clearly the fact that roles are an important feature in equality is relevant in Haselton’s theory but not Landsburg’s. Both positions are underargued.

      • Alex Perrone

        Ah man, unfortunately after more thought (I think the post is so good that it’s worth it), I realized that this critique 3 had just a funny wording for me. I took “obviously relevant in your theory T” to mean that it is obvious that it is relevant in your theory. For Haselton, it is obvious that roles are relevant, therefore critique 3 is not really devastating.

        But what critique 3 is getting at I think is better shown by this re-wording:

        “3. But it is not obvious why feature R is relevant in your theory T.”

  • Alex Perrone

    The problem with this example of is that the employer/employee relationship is supposed to be not obviously relevant, when of course the status quo is the opposite. As Landburg stated but didn’t argue for, the fact that they are playing different roles is an irrelevant circumstance. Haselton counters that it’s relevant, so in a way they aren’t disagreeing fundamentally on the principle of equality they both stand by. They simply apply their standards of relevance differently.

  • keddaw

    Interesting post, as an economist and (mostly) libertarian it seems to me people should be able to discriminate in both hiring and job selection processes. In a perfect world both these choices would show long term problems and employees using this strategy would be rewarded less and companies with the strategy would be out-competed.

    In the real world there are huge differences in the two situations. Forcing a company to choose the best candidate is somewhat different to forcing someone to work against their will – we call that slavery.

    There is another real-world problem; it is infinitely harder to prove why a candidate rejected a job offer than to prove systematic racism (or any -ism) by an employer simply due to the number of jobs a company offers (and number of applicants) compared to the number of jobs an individual applies for and is offered. And also the willingness of the wronged party to follow it up.

    So should an employer be given the right to sue a (non-)employee for discrimination? If the (non-)employee can sue the employer then yes.

  • Aaron

    There’s an obvious, massive asymmetry. An employee takes orders from his employer. Yes, the employer has to “deal with” his employees, but there’s a huge difference between giving orders and taking them. If non-discrimination applied to choice of employers, then the government is telling private parties to spend large chunks of their time obeying certain particular other private parties (namely, the employers whom they would choose apart from their discriminatory practices). Bad in itself, and a dangerous precedent.

  • Pingback: You said it better than my years of attempts | Entitled to an Opinion