On Thursday, I blogged about what struck me as a very troubling decision by the Equal Employment Opportunity Commission. An employee complained that a co-worker repeatedly wore a “Don’t Tread on Me” cap to work, and that this was “racially offensive to African Americans” because “the flag was designed by Christopher Gadsden, a ‘slave trader & owner of slaves.'”

The employer rejected the complaint, but the EEOC ordered the complaint reinstated, because “whatever the historic origins and meaning of the symbol, it also has since been sometimes interpreted to convey racially tinged messages in some contexts.” (The EEOC gave two examples, one of “assailants with connections to white supremacist groups drap[ing] the bodies of two murdered police officers with the Gadsden flag during their Las Vegas, Nevada shooting spree” and “African-American New Haven firefighters complained about the presence of the Gadsden flag in the workplace on the basis that the symbol was racially insensitive.”) Therefore, the EEOC concluded, “Complainant’s claim must be investigated to determine the specific context in which [the co-worker] displayed the symbol in the workplace.” The EEOC mentioned no allegation that the co-worker personally insulted the complainant; no such insults would thus be part of the “context.” Presumably the employer would need to investigate whether some other statements in the workplace or in the news gave the flag a supposedly “racially tinged” message.

This, I argued, is very dangerous to free speech. Whatever power employers may have to restrict speech in their own workplaces, the government — in the form of the EEOC — can’t use its coercive power to threaten employers for allowing political speech, even if the “context” suggested that the speech is “racially tinged” or “racially insensitive.” Such content-based, viewpoint-based restrictions imposed by the government violate the First Amendment. (This is a brief summary of my post, and thus necessarily something of an oversimplification; for more, see the post itself.)

Harvard law professor and Bloomberg View columnist Noah Feldman then wrote a column Friday responding to me. Now I much respect Prof. Feldman, and often agree with his work (see, e.g., this post). And, to his credit, he acknowledges that the EEOC’s actions are potentially troubling. But his ultimate defense of the EEOC’s actions, I think, only helps show how dangerous some aspects of “hostile work environment harassment law” has become. Consider this passage:

There’s something troubling about a government inquiry into the meaning of speech. As Volokh correctly points out, federal regulation of workplace political speech has no obvious stopping point: What if, he asks, someone considers a pro-Donald Trump workplace symbol to be racial harassment?

The problem with this argument is that it proves too much. Any form of prohibited workplace harassment, whether based on race or sex, can be mixed with a political message. If someone says in the workplace that Hillary Clinton shouldn’t be president because women shouldn’t work full-time, that’s a political statement. Yet it could also be part of the pattern of sex discrimination in a hostile work environment.

According to Prof. Feldman, then, harassment law can require employers to suppress even speech criticizing a presidential candidate, if that speech expresses a sexist viewpoint. Indeed, nothing about Prof. Feldman’s analysis turns on Clinton being a candidate — if Clinton is elected, criticizing her actions as president on the grounds that her sex is causing her to act unsoundly would likewise help create a “hostile work environment.” Employers would thus be obligated to suppress such criticisms by their employees or face the risk of massive civil liability.

But wait: Prof. Feldman’s Hillary Clinton examples comes in a column that explains why the EEOC is right to order an investigation of a co-worker’s displaying a Gadsden flag, even though it’s not racist on its face — according to the column (and the EEOC), the employer has to “investigat[e] … the specific complaint’s context to see if workplace racism was in play.” This means that even if an employee complaints that a co-worker was criticizing Hillary Clinton without using overtly sexist arguments, but the employee perceived the underlying message as sexist, the employer would then have to investigate the “context to see if workplace [sexism] was in play.”

And of course people have argued that a vast range of criticisms of Hillary Clinton are sexist: That “She doesn’t connect. She isn’t likable. She doesn’t inspire. She seems shrill. ‘She shouts.'” That she wears a $12,000 jacket. That her success is due to her marriage to Bill Clinton. That she is “polarizing, calculating, disingenuous, insincere, ambitious, inevitable, entitled, over confident,” or “secretive.” The list could go on.

So, by the logic Prof. Feldman’s defense of the EEOC’s Gadsden flag decision, any time someone complains to an employer that a co-worker repeatedly condemned candidate Hillary Clinton — or, if she is elected, President Hillary Clinton — as “calculating,” “insincere,” “secretive,” or not “likable,” the employer has to investigate whether, in context, a reasonable person (or, more likely, a future juror) could perceive the statement as sexist. If you are a reasonable employer, you’ll get the message loud and clear: Saying things that criticize Hillary, and that some might (rightly or wrongly) perceive as sexist, exposes you to the risk of a massive damages award. And the consequence will be that you will just order your employees to stop saying such things, even if you’d rather be more politically tolerant, because of the risk of government-imposed liability.

Pro-Trump statements? Too dangerous for employers to allow, so long as someone perceives them as sending an implicitly anti-Hispanic or anti-Muslim message. (Prof. Feldman doesn’t deny that such statements could require employer investigations, even if they are just “Vote Trump”; his response to “What if, [Volokh] asks, someone considers a pro-Donald Trump workplace symbol to be racial harassment?” is ultimately “Speech that’s constitutionally protected elsewhere is restricted at work,” and by this he means restricted by the government’s coercive power and not just by private employer choice.) Anti-Clinton statements? Too dangerous for employers to allow. How convenient….

Ultimately, Prof. Feldman’s argument comes down to this:

The theory of our laws against harassment is that the workplace is different from the public square. Speech that’s constitutionally protected elsewhere is restricted at work precisely because we have a countervailing social goal of creating a discrimination-free workplace.

What makes this constitutional is a subtle question, but the simplest answer is that we are prohibiting a course of conduct at work, namely discrimination. The harassing words are prohibited because they are the mechanism whereby discrimination occurs. Because we’re in a controlled environment, not the public square, we allow the goal of anti-discrimination to outweigh the cost to self-expression.

Workplace anti-discrimination law is an experiment, of course. But so far, it seems to be a worthwhile one. Investigating potentially racist symbols at work isn’t a small price to pay — but it may be a necessary price.

I’ve responded to such arguments in detail here, and, in particular here (as well as in other places). But for now, let me offer two big picture observations:

1. Work is where most of us spend a third of our waking hours. For most of us, it’s where we spend the most time taking to people outside our own family. Letting the federal government suppress “potentially” sexist, racist, or anti-religious speech throughout the nation’s workplaces, public and private — essentially, a nationwide speech code — is a vast interference with free discussion, much greater than the many far narrower viewpoint-based speech restrictions that the Supreme Court has rightly struck down.

2. Almost every place is someone’s workplace. The Gadsden flag case arose in a post office, but it could also arise in a newspaper office, at a university, in a restaurant, in a bookstore, or in a vast range of other places. If employers have to investigate and potentially suppress Gadsden flags or anti-Hillary statements when employees are offended, then they have to likewise investigate and limit speech when university staff are offended by statements within the university, when waiters are offended by T-shirts worn by regular patrons or jokes said in the waiters’ presence, when newspaper employees are offended by what is printed in the newspaper (and what is said about such material during the editing process), and more.

I think that workplace harassment law may be properly applied to physical abuse, to threats, and likely to unwanted speech said to a particular person — the bulk of what harassment law has traditionally suppressed. But Prof. Feldman and the EEOC think it can also be applied to political statements said to potentially willing co-workers, so long as some co-worker is offended, and so long as the “context” suggests that “racism,” sexism, or religious hostility is “in play.” And that includes, in Prof. Feldman’s view, criticism of presidential candidates, and presumably of the president him- or herself. That sort of government-coerced, viewpoint-based speech suppression, I believe, cannot be constitutional.