The idea that AI is gender neutral is complete BS—as long as the programmers are biased, so too is their creation
Roff 16 (Heather, Gendering a Warbot, International Feminist Journal of Politics, 18:1, 1-18, DOI: 10.1080/14616742.2015.1094246)
The base knowledge provided and the process by which it is accessed is key. What “facts” or pieces of information go into the database will affect the actions of the machine. If a data point or “piece” is not present, there is no opportunity for it to get there. An AI might be able to rearrange existing bits of information to combine them into new insights, but it will never generate something outside of its existing set of information structures. There is a gatekeeping function present in the construction of all symbolic representations. As Adam (1998) notes, “A formal system, in other words a computer system, is a political choice to maintain existing power structures” (108).16 To bring this back to gendered warbots, if programmers decide that “child” as an object or representation can never be paired with “combatant,” then the machine will not be capable of identifying child soldiers. Moreover, what constitutes “child” will have to be represented in some way. The process by which such “knowledge” is represented is thus crucial. As Adam (1998) explains, the problem about deciding “what sort of objects there are in the world” becomes another problem about knowledge about “facts,” and how a knowledge base about “facts” is recreated within a system of purely propositional knowledge” (39).17 The purely propositional knowledge approach is open to all of the feminist critiques from feminist epistemology, in particular, that there is some sort of Truth about which the facts must correspond and that knowledge derived from skills (such as a “knowing how”) is not real knowledge (Adam 1998, 114). Warfighting is as much an art as it is a science, and so prohibiting this kind of embodied or practical knowledge would appear to undermine the ability of a warbot to fight effectively. The most pressing problem for symbolic systems, however, comes from the direction of analogical reasoning. Faced with the certainty that warbots will encounter complex situations, which they cannot have prior knowledge of, they will require some mechanism to abstract the situation to one about which they do have some higher order representation. This is done by way of analogy. A program will have to break down the situation into smaller objects that it can represent, and then it will search its database for something that is analogous where the concepts are equivalent or appropriate. To find such an abstract analogy, the AI uses stereotypes (Adam 1998, 39).18 Programmers may not readily admit to this, claiming that “the important thing is to break the overall experience reflected in the memory into composable regularities” (Storrs Hall 2007, 221). “Composable regularities” might be that all hills are convex physical structures, but they might also be that “all women speak with a high voice” or “all combatants are men between 18 and 35.” Reasoning by analogy is a heuristic that humans employ when they are being cognitive misers and is not appropriate for complex situations requiring critical thought.19 This brings us full circle to the problem of which objects and concepts get represented in the dataset. If the programmers, who are predominantly white, male, upper-middle class and academic, equip an AI with these perspectives, then it is not difficult to see how gendered constructions will reproduce themselves (AAUW 2015). A particular masculinity might be reproduced, or a particular view of femininity may not be modeled at all (Adam 1998, 23). In either case, the AI is clearly not “gender neutral.”