When thinking about what it is to be biased, people tend to think of someone living in the backwoods, brooding over how “They took our jobs,” and cherry-picking statistics to self-validate their own prejudices against people of other colors, creeds, and backgrounds. Well, that’s stereotyping and shame on you for doing it.
Bias is overlaying one’s assumptions and simplifications on top of a complex and nuanced person, idea, system, or thing (1). Bias is part of the human condition, it’s how we function (2). Imagine trying to grasp every subtlety of any given situation at all times: it’s inefficient, impractical, and socially awkward. As functioning adults, and sometimes professionals, we are expected to just know things. Asking questions can feel imposing and embarrassing. That’s just how it is. We as humans are simply uncomfortable with acknowledging uncertainty. Bias is inherent and unending, and its minimization should always be pursued.
Bias is already a problem. There was already so much inherent bias in the way that individuals were living their lives that the law had to be changed (several times) in order to try to mitigate the effects of biases (3). However, as bad as it is, what’s the worst thing that can happen when an individual factors an implicit or even explicit bias into their decision? You think, “Wow, what an asshole.” What if that person is representative of, say, a particular restaurant in a community? You avoid that restaurant, and maybe you have a bad time the first and only visit you make to the restaurant. What if the person is representative of an entire town, state, or country? Suddenly the problem is no longer a negligible and easily avoided nuisance.
The problem with algorithmic bias is the difficulty in detecting it and its cold scalability (4). Even those who actively challenge their own biases can accidentally implement their own biases, and when you’re dealing with products that can be downloaded at the touch of a button and delivered to millions of people instantly, suddenly the scale of that minor problem becomes immeasurable. The problematic program scoops up data and spits it out like pulp from a mill. But despite all best intentions, we’re all subject to the law.
There are two main problems of law with bias. One problem is a priori and one is a posteriori. The experience of believing in the basic essence of a thing being universal to the plurality of instances of that sort of thing requires no applicable experience for the negative implications to be apparent — if one is operating on biases when approaching a person or situation, one is missing the richness of the entirety of the situation or person’s character. If one is experiencing a bias, one has already diminished the fullness of an experience. The a posteriori problem follows from the search for the a priori problem. The a posteriori problem is one of direct impact on the subject of the bias, as well as the indirect effects which are far more difficult to define. The indirect effect is the ripple effect, the thumb on the scale. The direct impact of bias is the imbalance created by the effect on the subject, the indirect impact is the affirmation of the initial bias.
This paper took a theoretical look at the legal personhood of AI. It was written in collaboration with Ritika Gopal, Associate Counsel at SXSW. You can find her at ritikagopal.com.
Technology is being sophisticated at incredible rates. There are people who have lived through the invention of flight, landing on the moon, iPhones, and DeepMind.[i] Technology has moved on from tools of human advancement to full-on integration with the human body.[ii] The potential for body modifications to surpass the abilities of the natural human state are here.[iii] Likewise, artificial intelligence is advancing at an incredible rate[iv] and is getting closer to modeling the human learning process.[v] There has not been much thought given to whether or how a conglomeration of these advancing technologies would accumulate rights, or if rights can already be conferred under existing legal theories (largely because of the unanticipated rate of development). This memo explores legal theories establishing possible rights of robots.