Attorney Sues Website Over His Online Rating

An anonymous reader writes "The Seattle Post-Intelligencer is reporting that a local attorney is suing legal startup Avvo over a rating that was algorithmically assigned. The story covers the controversy of computers grading humans. 'Browne, who has participated in a number of high-profile cases in the state, including the defense of arsonist Martin Pang, said in an interview that Avvo is being irresponsible with the ratings and called them a fraud. And he questioned why Supreme Court justices and prominent lawyers score so low. Three other attorneys interviewed by the P-I also expressed doubts about the rating system, while News.com reported that the site "seemed to be riddled with bizarre errors."' Such practices are not new: the New York Times earlier this year reported on Google using algorithms to determine applicant suitability. But what happens when you don't like the result? Can a computer program be considered defamatory?"

Comments

It's an input problem

[MightyYar]

There's nothing wrong with computer rankings, but they inputs are very, very important. We shouldn't pretend that they are different somehow from human rankings, since humans still carefully select the inputs.

An example that most here can relate to is the US News and World Report college rankings. It's a whole other topic in itself, but suffice it to say that there is a lot of discussion about their inputs and how it has influenced the way colleges operate. Most colleges try to get many small donations instead of a few big ones, because the rankings weigh number of donors more heavily than total amount donated. They encourage many, many applications from just about anyone because they get ranked based on the number of applications that they reject.

Once people learn what the inputs are, they just game the system.

Re:It's an input problem

[Dymus]

And it isn't terribly difficult to determine what the inputs are. I am ranked at 8.1 (only having been licensed for two years) which is higher than all but one of the attorneys practicing in the same field as me at my firm (ranging between 25 and 6 years of practice). After reviewing all of our AVVO pages, it appears my rating is boosted significantly because AVVO discovered I received two CALI awards in law school. Anyone in the business knows these awards are more or less meaningless in practice (especially since the two I got have nothing to do with my practice area), but they appear to have a significant impact on my score. Based on our informal analysis, it looks like you start with a baseline of around 5, get a boost of about 1 for every award they found for you, and reduced a ton if you have any bar disciplinary actions pending. I played around with it by updating my profile with organizational memberships and even got endorsed by an attorney friend of mine and it had no impact on the score. The key seems to be how much the AVVO system can find on you on its own, without you updating it. So more important than what the inputs are may be determining where AVVO is farming its data from. Overall, it seems to be that the criticism the ranking system is receiving is pretty valid.

Re:My Opinion?

[fyngyrz]

I can think of a few reasons why supreme court judges would be legitimately rated low. Support for blatant ex post facto laws; commerce clause inversion; allowing states to take people's property; allowing wiretapping before a warrant is issued (FISA); Allowing congress to blackmail the states by withholding highway (and other) funds; allowing the right to bear arms to be infringed upon; allowing government support of Christianity; outright ignoring the 10th amendment (see earlier reference to blackmail); cowardly and un-statesmanlike refusal to hear critical cases of government malfeasance (like Robert Newdow's); allowing the state to infringe upon the liberties of the citizens (drug war, censorship, marriage, sexuality, unreasonable copyright and patent terms and mechanisms; allowing the feds to step beyond the enumerated powers without requiring a constitutional convention; restricting freedom of speech (free speech zones, censorship, funeral zones, etc.)... that's all just off the top of my head.

Yeah, I could definitely see low supreme court ratings.

Re:"Can a computer program be considered derogator

[dkf]

switch (person.getSkinColor()) { ... } ?

It really depends on what you're doing with it. For example, if the program is recommending what hue of skin cream (or made-to-measure suit, or something else like that) will work best with your colouration, then that program that requests that info for such purposes can hardly be inherently defamatory. On the other hand, it is most certainly possible to use a program to defame someone: it's not that hard to conceive of ways to do it (e.g. by spamming libellous statements all over the place.) But all this is programs, not algorithms; I've no idea what a defamatory algorithm might look like, and I don't think it is possible to make mathematics work that way.

As an aside, I really doubt that skin colour is going to be expressible as a simple enumeration or integer, so you won't be able to switch on it...

bad program but not defamatory

[belmolis]

It sounds like the rating algorithm isn't very good, but I don't see how this guy can win a suit for defamation. In US law, only false claims of FACT are actionable. If the web site stated that a lawyer had received a reprimand from the bar association when he had not, that would be defamatory. If it said he had cheated a client or bribed a juror and he had not, that would be defamatory. But saying: "This guy is a jackass" or "This guy is a poor lawyer" is not actionable because these are opinions.

I'm not sure what can be made of the use of a poor algorithm. If they disclose the algorithm and say "Here is what we get when we plug in the data we have", so long as the data is accurate and they apply the algorithm correctly, they aren't making any false claims of fact. Ethically, it seems like there should be a penalty if they persist in using an algorithm that demonstrably does not produce output that is reasonably related to what people generally take to be valid measures of lawyer quality and if they deceive people into thinking that it is valid, but I'm not sure how this can be addressed legally. I think you'd have to argue that there is an objective definition of lawyer quality of which the algorithm gives a false view. I don't know if defamation has ever been proven on such a basis.

Re:Just another tool.

[Mistlefoot]

And if my algorithm says that if your first and last name start with P you must be a PedoPhile can I call you one and hide behind math?

I think not.