(Fewer) Rights for Algorithms?
posted by Frank Pasquale
Yale Law Prof. Ian Ayres’s Super Crunchers celebrates the new era of data-driven decisionmaking. A NYT piece provides a nice introduction to these technologies:
[W]hen so much data is processed so rapidly, the effect is oracular and almost opaque. Even with a peek at the cybernetic trade secrets, you probably couldn’t unwind the computations. As you sit with your eHarmony spouse watching the movies Netflix prescribes, you might as well be an avatar in Second Life. You have been absorbed into the operating system. . . . [W]hen executives at MySpace told of new algorithms that will mine the information on users’ personal pages and summon targeted ads, the news hardly caused a stir. The idea of automating what used to be called judgment has gone from radical to commonplace.
Jeff Lipshaw asks, in response: “is it possible to program [such automation] so powerfully that it replicates all possible human (i.e. brain) programming?” And Larry Solum helpfully brings up a 1990 article he wrote on the implications of such questions for law: “Could an artificial intelligence become a legal person?”
Though such a possibility might seem a long way off, it is embedded in some recent legal arguments of Google. Google is perhaps the world’s premier example of “automating judgment;” its engineers are constantly thinking of new ways to order information in response to search queries. According to one of its court filings, “Google takes extraordinary measures to protect its trade secrets and confidential commercial information.” While resisting any efforts to “peak under the hood” of its search processes, Google also has been claiming that whatever results they come up with should be protected under the First Amendment. So one of the questions posed by Solum has nearly come to a pass: Google is seeking constitutional protection for what (it assures us) is an entirely automated process. Should it get it?
Let’s follow a thought experiment parallel to Solum’s inquiry to begin thinking about the issue. Imagine that, instead of Google, Inc., a robot that ordered search results claimed that its outputs were First Amendment protected speech. It will (be programmed to) argue that “it is a person, and that it is therefore entitled to certain constitutional rights.” Though I am abstracting from an extraordinarily rich and complex paper, I find it helpful to note here Solum’s immediate response to that possibility:
Should the law grant constitutional rights to AIs that have intellectual capacities like those of humans? The answer may turn out to vary with the nature of the constitutional right and our understanding of the underlying justification for the right. Take, for example, the right to freedom of speech, and assume that the justification for this right is a utilitarian version of the marketplace of ideas theory. These assumptions make the case for granting freedom of speech to AIs relatively simple, at least in theory. Granting AIs freedom of speech might have the best consequences for humans, because this action would promote the production of useful information. But assuming a different justification for the freedom of speech can make the issue more complex. If we assume that the justification for freedom of speech is to protect the autonomy of speakers, for example, then we must answer the question whether AIs can be autonomous.
Solum also considers a number of objections to granting the AI itself rights; for example:
[T]he “paranoid anthropocentric” argument [runs:] “AIs might turn out to be smarter than we humans. They might be effectively immortal. If we grant them the status of legal persons, they might take over.”
***
The second objection, that AIs lack some critical element of personhood, is really a series of related points: AIs would lack feelings, consciousness, and so forth. The form of the objection, for the most part, is as follows. First, quality X is essential for personhood. Second, no AI could possess X. Third, the fact that a computer could produce behavior we identify with X demonstrates only that the computer can simulate X, but simulation of a thing is not the thing itself. X is that certain something–a soul, consciousness, intentionality, desires, interests–that demarcates humans as persons. Call this argument, in its various forms, the “missing something” argument.
***
Finally, the third objection to constitutional personhood for AIs is that, as artifacts, AIs should never be more than the property of their makers. Put differently, the objection is that artificial intelligences, even if persons, are natural slaves.
The third objection points us in the direction of a more immediate resolution of the problem here: in our case the algorithm is a tool of an existing corporate entity, Google. But should the fact that Google’s results are automated lead them to get less protection than, say, a social search engine that ordered the web? I think so, for reasons largely derived from the first two objections to “rights for AI’s” mentioned above, and also because of the secrecy of the Google search process.
In terms of the AI argument, one might claim that Google is much closer to a data provider than, say, a newspaper. The latter actually expresses a point of view on what the news is; the former merely aggregates information. This difference has consequences for law.
Data providers like consumer reporting companies can be held more accountable for what they say than a newspaper. If I have a dispute with a newspaper over whether they’ve portrayed me accurately, I’m probably going to have to sue for defamation in order to settle things. But according to an FTC website, “If an investigation doesn’t resolve your dispute with the consumer reporting company, you can ask that a statement of the dispute be included in your file and in future reports. You also can ask the consumer reporting company to provide your statement to anyone who received a copy of your report in the recent past.” Moreover, ” only authorized individuals such as potential lenders, employers, insurance underwriters or landlords may access your report, and only if they intend to do business with you.” Finally, in case of disputes, “you’re entitled to add a written statement (100 words or less) explaining your view of the mistake.”
Why might we want to extend this type of distinction to the search world (and, indeed, strengthen consumer protections vis a vis “black box” data aggregators like ratings agencies and FICO scorers)? I think that there is something deeply troubling about unaccountable power–about a system that can simply spit out some life-changing result without giving a full explanation for it. Suspicion about FICO scores has led some states to prohibit their use in insurance rating, just as Finland has prevented employers from using Google results in evaluating potential applicants. Full First Amendment protection should be reserved for accountable, attributable speech–not the data processing systems that are increasingly powerful arbiters of taste, authority, and creditworthiness.
October 4, 2007 at 8:52 am
Posted in: Google & Search Engines
Print This Post
Responses (7)
James Grimmelmann - October 4, 2007 at 11:22 am
Your point about unaccountability applies equally to the social search engine. There is no one within the social system who really takes responsibility for the results.
As for personhood, the Google algorithm falls well short of the complexity and behavior that would lead us to regard it as being a person in a meaningful philosophical or useful sense. But an AI that produced search results and were in other respects sophisticated enough to be a proper subject of personhood arguments — I think you might find that it looked a lot better than the Google algorithm does, in terms of its speechiness.
Don’t be misled by automation versus manual decision-making as a categorial matter. Algorithms express the values and choices their programmers pour into them. There is speech here, and it’s valuable. The kinds of arguments you want to make hinge on secrecy and on the massive power that search engine rankings wield.
Bruce Boyden - October 4, 2007 at 11:49 am
I’m not sure I understand the relevance of Solum’s article here — although it’s fascinating. (As are the various science fiction stories that deal with this question. While off on that tangent, I should note that I’ve always viewed 2001: A Space Odyssey as in part a cautionary tale about why you shouldn’t make your operating system sentient. But I digress.) Just because Google’s algorithm is not itself a person doesn’t mean that Google doesn’t have any rights in it.
There is an interesting question here though about who can claim rights in works/results/etc. produced by programs. I.e., stuff that is one level removed from the original programmer. Obviously the programmer can claim speech and IP rights in the program itself. But can he or she claim rights in what the program produces? E.g., musical works spontaneously generated by the program? The answer to that seems to get into the weeds rather quickly.
Frank - October 4, 2007 at 7:50 pm
To Bruce: Yes, Hal is pretty scary to me. . . . as, in his own way, is the robot-kid in Kubrick’s AI! Kubrick is asking really tough questions about the future evolution of humanity.
And I concede that the line I am trying to draw between “algorithm-driven results” and corporation or person-driven results may be too weak a hook on which to hang any persuasive legal distinctions. (forgive the mixed metaphor!) But to build on James’s point, perhaps my key argument here is that, if we don’t know how a decision is made, we might want to limit the range of its effects.
James: I have to reread the relevant cases on software as speech (I think there is a Robert Post article on the topic I really need to get my hands on). But I guess I would fall back on the idea of levels of First Amendment protection. It seems to me that Google is fundamentally like a map of the web, rather than an expression of a point of view. That certainly is the way they present themselves; see, e.g., this comment I gleaned from Google’s SEC Reg. Statement (hat tip to Lastowka’s latest):
“Google users trust our systems to help them with important decisions: medical, financial and many others. Our search results are the best we know how to produce. They are unbiased and objective, and we do not accept payment for them or for inclusion or more frequent updating.”
And Bracha and I cite some cases in our FSC article where mapmakers are denied First Amendment protection for their wares. I would not deny Google protection in toto because there are a lot less “degrees of freedom” in webmapping than in earthmapping. But I also think that there needs to be accountability in the former profession, especially if (as I believe) one company is likely to be doing an ever-increasing share of the production of the basic results everyone else uses.
Now, as for “social search,” perhaps I should have said “human powered search,” because I was thinking of Mahalo. For example, I think there is an actual editor in charge of a page like this:
http://www.mahalo.com/Trip_to_costa_rica
Of course, some say that competitors like Mahalo make Google’s dominance less noteworthy than I think it is. But this page suggests that Mahalo is less a competitor than a coadjutor of Google:
http://mahalo.com/Special:Search?search=Trip+to+Puerto+Rico
Frank - October 4, 2007 at 7:55 pm
Mahalo follow up: yes, at the bottom of the page it says:
This search result page was written by:
Jonathan
Mahalo’s goal is to hand-write the top 20,000 search terms.
You can help by recommending your favorite links.
Vickie Pynchon - October 5, 2007 at 1:00 am
fewer (not less)
sorry; can’t help pulling out the blue pencil from time to time . . .
Frank - October 5, 2007 at 10:06 am
Duly corrected, Vickie!
Samir Chopra - October 5, 2007 at 4:13 pm
Frank,
You say “if we don’t know how a decision is made, we might want to limit the range of its effects.” in the context of trying to draw a line “between algorithm-driven results and corporation or person-driven results”. I’m not sure there is such a principled distinction to be drawn. First, we assume too much knowledge on our part about the reasons why humans come to certain decisions. Very often human beings confabulate reasons for the decisions they make (there is a large psychological literature on this). And secondly, our own decisions are to a certain extent programmed, by societal expectations, education, culture, upbringing, language, gender, nationality and what have you. Where one kind of decision changes from the “purely algorithmic” to the “decidedly human” is not clear.
Later, you go on to say, “It seems to me that Google is fundamentally like a map of the web, rather than an expression of a point of view.” But maps can very often to be expressions of point of view, reflecting the biases or the interests or values of the map maker in what they choose to depict, highlight, ignore or selectively crop. Google’s map-making, for better or worse, generates a contoured map of the Internet landscape, pointing out routes we should take on our travels and often, suggesting what we should buy along the way!
Incidentally, in the software as speech cases, are you thinking of Bernstein v. US DOJ 1999? Or the DeCSS case?
Leave a Reply