Roommate.com reversed
The Ninth Circuit reversed the lower court in Fair Housing Council v. Roommate.com, limiting Section 230. Section 230 is the federal statute that immunizes online services providers for their users’ content. Most courts have construed Section 230 broadly, protecting ISPs against all sorts of liability. Intellectual property is (surprise) specifically exempted.
In this case, the Fair Housing Council sued Roommate.com for violations of the Fair Housing Act, which prohibits housing advertisements that state discriminatory preferences. The lower court held that Roommate.com, as an online provider, was immunized by Section 230 for postings on its service that allegedly violate the Fair Housing Act’s discriminatory housing ads prohibition.
The Ninth Circuit’s opinion, written by J. Kozinski, distinguished Roommate.com from a protected service based on three main features:
* Roommate.com shaped its users’ input with a form questionnaire that specified gender, sexual orientation, and with-children choices;
* Roommate.com’s search mechanism which has the same pulldown search options (gender, sexual orientation, and with-children);
* Roommate.com emails people with listings based on profile preferences (gender, sexual orientation, and with-children)
J. Reinhardt would have also held Roommate.com liable for the discriminatory content posted by users in the comments section, but the majority held that that information, which was not solicited by nor built into the search engine, was not Roommate.com’s responsibility.
J. Ikuta dissented from liability, and would have followed Carafano finding broad immunity even on the form submission.
… So what does it mean? My quick take is this:
First, readers will probably be familiar with these issues from the Craigslist case, and concerned about the 7th Circuit appeal. In my opinion, this case can be reconciled with the decision in the Craigslist case (ND Ill, a 7th Circuit court). This decision expressly held that the “no parameters” forms that are just open-ended do not create liability, and Craigslist.com’s housing forms are primarily open-ended: address, rent, cats/dogs. That decision is on appeal to the 7th Circuit. Should the 7th Circuit elect to address this aspect of the decision, that’s a fairly clear distinction to make. The 7th Circuit need not address this, though, because both the search & the email features that Kozinski brought up really hinge on the structured data input, which is the only real point of distinction between a “content provider” and a “content service”.
Second, despite Kozinski’s dancing, the case will be harder to reconcile with Carafano v. Metrosplash.com, a 2003 9th Circuit case that considered defamation & privacy liability for an online personals service that set up forms for publishing user input. (Carafano, 339 F.3d 1119.) In particular, it adds some weight and significance to the distinction between an “ICP” and an “ICS” (information content provider and interactive content service, respectively).
Third, the decision’s major point of distinction between Carafano and this case was that the ISP established policies. So establishing policies that reiterate the law will be key for ISPs in the wake of this decision. Kozinski stressed that the service in Carafano did not solicit the problematic information and in fact expressly forbade some aspects of it. So, under this decision, establishing policies that reiterate the law will go some way toward protecting an ISP. While this isn’t the worst outcome for a speech-related law, it seems (to me) to be a waste of time, and I’d point out that it burdens ISPs with educating their users about the law. This sort of burden is, to my mind, inconsistent with notions of ISPs as “utilities”, and also inconsistent with the broad, unfettered access to communications that the First Amendment contemplates.
Fourth, I imagine that this case will be used by all who seek to limit Section 230’s broad immunity for ISPs. Whether the case constitutes a high-water mark for limitations on that liability immunity, or merely a beachhead, remains to be seen.
update: Eric Goldman analyzes the decision at greater length, and you can get the flavor from the title “Ninth Circuit screws up 47 USC 230″.
PS - More on the FHA: I always think about, and rarely remember to point out, that one of the contested categories is sexual orientation, which isn’t included in the Fair Housing Act, although various state anti-discrimination codes do include it. Some people have also been confused about the roommate exception, which permits discrimination by roommates for shared-housing situations. If you lease a room in your apartment or house, for instance, you can discriminate on religion, etc. (Although shared-housing people can discriminate, advertising those discriminatory preferences is not protected.) So, generally, commercial landlords may not discriminate on race, sex, familial status, etc. (but they can discriminate on sexual orientation); shared-housing lessors may discriminate; advertising discriminatory preferences is prohibited, period. (But, again, advertising preferences against or for sexuality would be okay under the FHA.)
algorithmically similar posts:
» roommates.com: no 230 safe harbor for discriminatory housing ads, 2008-04-03 (score:50)» A failure of the public interest tech law community, 2006-11-19 (score:30)
» gay marriage & Equal Protection jurisprudence, 2008-05-16 (score:19)
» Section 108 report released, 2008-03-28 (score:18)
May 15th, 2007 at 3:59 pm
Okay, so I’m just a little old civil rights lawyer, rather than a cyberlaw stud, but I don’t see what’s so terrible about this opinion. Isn’t Judge Kozinski totally right that a contrary ruling would mean that the website could ask users “Do you have any racial preferences?” and be immune from answering for that under the Fair Housing Act? That seems to be the clear result of the rule Prof. Goldman seems to prefer: “So long as any third party ICP was responsible in part for the content’s development (even 0.1%), the ICS isn’t liable for it.” Leave aside whether this is what Members of Congress thought they were doing — the statutory text certainly doesn’t require such broad immunity, and Prof. Goldman seems to admit that the text also bears the reading Judges Kozinski and Reinhardt put on it. Just on first principles, why is such broad immunity such a good thing? Newspapers fought and lost this battle in the early days of the Fair Housing Act, and rightly so. What makes a business like Roommates.com any different?
May 15th, 2007 at 4:29 pm
Let me make a couple of specific decision points.
(1) In terms of trying to lay out a line between 230 and the FHA, the ICP/ICS line may not be a bad one, if it’s drawn correctly. (I’m not sure this one is.)
(2) One problem with this case is that it’s quite different from virtually all the other Section 230 cases. Just in terms of counseling clients and telling ISPs what to do, that creates a problem.
(3) Another problem with this case is that much of what Kozinski does is a type of handwaving. He lists email & search of the field-delimited results, but he’s not completely and straightforwardly clear about email & search liability relating to the field-delimitations (gender, family structure, sexuality). So that lets Reinhardt lay out his version of the test, and now there are two competing versions in one opinion. So if email & search are only a problem because of the field limitations with proscribed fields (as I read Kozinski’s opinion), then they are being used just to lend a little rhetorical weight to the opinion. But that’s dangerous because not everybody gets the technical language, so some people (like Reinhardt) are going to read it much broader to include a wide variety of search & email functions. That’s really a problem, as Eric Goldman points out.
Then, “why is such broad immunity such a good thing?” There are lots of responses to this, but without rehashing the newspaper / FHA battle, let me hit just a few of the salient points that distinguish print from online media:
(1) Imposing screens on online communications is different from imposing screens on print communications. No matter how busy the Village Voice is, there are still a finite number of print ads to be screened. For an online Internet service, there are orders of magnitude more content. To do manual filtering would destroy all the benefits of the Internet. To do automatic filtering would require reliance of filtering programs that have proven, time and again, to be both over- and under-inclusive in any domain they operate.
(2) One of the primary rationales behind regulation of print media has always been that speech is free to people who own printing presses. In other words, that there is a significant burden placed on people to try to remedy unfair, libelous, or wrongful speech of various sorts in newspapers, broadcast TV, movies, etc. On the Internet, the ability of individuals to counter undesirable speech of all sorts, without resort to government regulation, is vastly increased.
(3) Another major rationale behind regulation of print media is that print media has a kind of permanence that spoken media does not. Hence, the distinctions between libel (print or fixed media defamation) and slander (spoken word defamation). The Internet is really somewhere between those two. Internet content can be easily edited or removed, like spoken word content, and can disappear forever (libraries and archives will testify to that). Certain kinds of content (e.g., porn videos of celebrities) have a kind of persistence and will be archived in a distributed manner by a wide variety of people, thus ensuring ongoing access for a while, but it’s unreliable, and only applicable to some kinds of content. (Classified ads are not usually made available on P2P sites, for instance.)
May 15th, 2007 at 5:31 pm
All lawyers advising clients want clear rules, and often appellate courts can’t come up with them without first working through a series of precedents. It’s life. And I don’t see why the arguments for broad immunity really work. (1) seems misplaced for two reasons: First, as I read Judge Kozinski’s position, Roommates.com isn’t going to be put at risk of doing any kind of filtering; it’s all about how they structure the questions. Second, the difficulty of filtering might well be relevant to the question of liability under whatever substantive law the plaintiff is invoking, but sometimes it’s going to be easier than others, so a broad rule of immunity — regardless of the cyber-context or the cause of action — doesn’t make sense. As to (2) and (3), they misperceive the reason behind regulating discriminatory advertisements. The point isn’t that they’re hate speech or group libel (for which an opportunity to respond and permanency might be relevant). The point is that they are either part of an illegal transaction or proposing an illegal transaction. Maybe it’s hard to filter these things out in a particular context, so a website shouldn’t be strictly liable if one happens to make its way on the site, but that’s a far thing from the blanket immunity we’re talking about here.
May 15th, 2007 at 8:39 pm
You asked about “first principles” and why is such broad immunity a good thing, and the reasons I gave are the reasons for a generally broad immunity for Internet speech. If we agree on that as a first principle (and it sounds like we may not), then one can ask whether it’s better to carve out exceptions (like the FHA rules) or to hold a firm line on the general principle.
Regarding counseling & clear rules — sure, it’s up to lawyers to come up with guidance based on various precedents. I’m fairly familiar with Carafano, though, and it will be tough to steer a course between Carafano and the holding in this case. I’m not sure it can be reasonably done, actually, and I’m not sure I buy J. Kozinski’s attempt to do it. In that case, most attorneys will probably advise their clients to go more conservative. On these facts, one might say “no biggie”; the only thing lost is search efficiency. (It’s more efficient to do searching of specific fields than to do keyword searching of a free-text field.) But not all matters will have facts like these, and no doubt in other situations there will be more lost than mere search efficiency. (Also, search efficiency might be a big deal for some clients and some types of resources.)
At any rate, on this case, so individuals can put whatever they want in the free-text fields which will be keyword-searched, but you can’t develop search forms for it. Does this outcome make sense? It seems like either an exercise in hypocrisy or an exercise in willful blindness to the way computer technology works or an exercise in stupidity. I’m not a big fan of court decisions that come down to minor distinctions in present-day technology.
May 16th, 2007 at 5:19 pm
I think that the proper advice is that: You should not rely on S. 230 immunity to the extent that the features that you offer are tailored do more than provide an opportunity to do bad things.
I see this as distinguishing between enabling law breaking and merely providing an opportunity.
I read this case a lot narrower than some, and I don’t read this case as saying that you cannot ever, under no circumstances, use structured text. I don’t even read it saying that you can’t use structured text which may, under some circumstances, be used to allow end users to do something illegal. Rather, in this case, it seems pretty clear that the tools were constructed in a way that offered little or no lawful purpose. Furthermore, it was not mere coincidence that certain combinations of the structured texts resulted in illegal grouping. I’ll leave it to the civil rights attorneys to defend whether this should be the case (though, I have my own doubts about whether I think that the FHA, as applied, makes good sense).
Finally, it’s easy to construct arguments against broadly immunizing websites that have structured text that users can create. I used this example on Volokh’s site: think Madlibs ™ with select violent verbs and select celebrity names which users can create and which are published to the front page (e.g. “Someone should [violent verb] [name].”). You probably wouldn’t want to immunize that behavior.
May 17th, 2007 at 3:05 pm
I think that another good point of distinction here will be the amount of control that Roommates.com exercised with the structured data. It required a user to disclose personal info on the protected characteristics, and then proceeded to filter not only the people who could see that user’s profile, but also who that user could see. There’s something much more active here than simply providing users with filtering tools. There’s something more than mere selection and editing going on here, and something more akin to an implementation of a discriminatory system, not merely a filtering tool.