1) The Use of AI and Statistical Methods in Legal Decision-Making
While many articles argue that the use of algorithmic assessment tools in parole-hearing is unwarranted, only a few works have explored the obligations of the state given that the state would use these tools in parole-hearing anyways. In this paper, I argue that such a state has a duty to use the same or a very similar AAT for the purpose of identifying the individuals who are less likely to be granted parole to offer them extra services. My argument is constituted of three stages. In the first stage, it is motivated that the state has a duty to accelerate the release of the incarcerated if it possible. I explain that while this duty obliges the state to provide extra support for all the individuals who anticipate parole-hearing, it gives the priority to the individuals who are less likely to be granted parole. However, the lack of a reliable method for identifying these individuals makes it impossible for the state to respond to the requirements of this priority (namely, to provide extra services emergently even when resources are scarce). In the next step, I explain that such a method indeed exists given that the outcome of parole-hearing is partly determined by the evaluation of an AAT. The individuals who are identified by the same or a very similar AAT are arguably less likely to be granted parole (at least, if everything remains the same); therefore, they should be offered extra resources that can reasonably reduce their risk scores. In the third stage, I take the last step to shift some of the burden of using AATs for parole-hearing from the so-called “high-risk” individuals to non-incarcerated citizens by arguing that the funding for providing extra services should be deducted from the resources that otherwise would have been spent on the non-incarcerated citizens. The problem would be how to divide the resources between the two groups. The solution relies on the idea that if two parties are to divide a cake fairly, the party that divides the cake into two unequal pieces should not receive the bigger piece. Similarly, the non-incarcerated citizens had a more substantial role in determining the situation in which one group would be cut off or not receive some resources, namely, by legitimizing the use of AATs for parole-hearing which triggered the duty to allocate the extra resources to “high-risk” individuals. Therefore, given that one party (either the incarcerated-individuals or the non-incarcerated ones) would end up with the less share of resources, it should not be the incarcerated-individuals who are deprived off the resources. In the meantime, if the arguments in the first and the second steps are sound, the “high-risk” individuals have the priority in receiving these resources.
How similar are the Precogs in Minority Report (Spielberg, 2002) and COMPAS in the modern legal system?
2) Vagueness and Non-literal Speech in Law
A term is vague if and only if there are cases – borderline cases – to which neither the term nor its negation is justifiably applied. There are vague terms in law. For instance, it is clear that the term “business establishment” in Unruh Civil Right Act applies to hotels and does not apply to houses, but it is not clear if it applies to Boy Scout of America or not. Many philosophers have tried to present an account to explain the implications of vagueness in law, both for law and for language. One of these accounts proposed by Roy Sorensen holds that judges necessarily lie when they face borderline cases. Sorensen explains that judges, like other language users, cannot justifiably apply a vague term or its negation to borderline cases; however, they have a professional obligation to choose either of the options. Resultantly, judges necessarily lie when they decide borderline cases, Sorensen says.
I argue that there is an assumption that Sorensen has taken uncontroversial which holds that the legal application of a term is permissible only if its ordinary application is allowed according to the rules of language. In contrast, I show that while judges do not apply vague terms to borderline cases (since it is unjustifiable as Sorensen explains), they use vague terms metaphorically. The main duty of judges is to request fitting legal consequences for involved parties. As a result, using a vague term for a borderline case just implies that judges intend the same consequences for the borderline case that they would intend for a clear case. This account not only refutes Sorensen’s counter-intuitive argument, but it also accounts for certain legal decisions in the history of modern legal system.
 See: Randall v. Orange Country Council (1998).
In addition, I am currently writing a review on The Nature and Value of Vagueness in the Law (Law and Practical Reason) by Hrafn Asgeirsson.
3) Value Sensitive Design: Towards Feasible Flexibility
Philosophy of Technology and Value Sensitive Design have gained noticeable attention in recent years, and they have become mature areas of philosophical studies. One of the pillars of Value Sensitive Design is the stress on the claim that technical artifacts and sociotechnical systems (hereinafter “technologies”) should promote the current values of society. To achieve this goal, engineers sometimes ought to design flexible technologies to facilitate promoting different values in different situations (for instance, for the situations that the values of society change), while they sometimes have to design inflexible technologies that only serve a singular value. However, in certain technologies always being flexible or always being inflexible is not desirable. We may need to temporarily take flexibility from a flexible technology or add flexibility into an inflexible technology. Thus, proposing a design approach that enables shifting between flexibility and inflexibility is advantageous. In this paper, I explain that the practice of computer programming has eye-opening intuitions for other engineers in this respect: Programmers have been embedding a feature into their designs that, considering the literature of the subject, can be called as adaptable flexibility, and I propose that the same feature should be considered in all engineering designs.
Computer programmers differentiate between two controlling cores that determine the tasks of technologies. One core is responsible for realizing the direct orders of users. For example, in Care Robots a controlling core is in charge of fulfilling daily needs of the patient. This core embeds flexibility into the technology. Meanwhile, another core is in charge of restricting flexibility if the predominant values are at stake. For instance, in care robots, the tasks of the first core are suspended and the second core takes control when the safety of the patient is in danger. This core thus realizes adaptable flexibility. In this paper, I argue that embedding two excluding controlling cores that are sensitive to different inputs – like in the case of Care Robots – is the key to adaptable flexibility. I propose that other engineers have to pay attention to this feature – if they already do not – as they should consider shifting between flexibility and inflexibility, while embedding adaptable flexibility into their designs is not necessarily desirable in all cases.
Presented in the Graduate Conference PIPR at Carleton University, Ottawa, on 1 Nov. 2019. Doi: 10.13140/RG.2.2.12208.00003.
Click here for the news on the SFU website.