User:Jemimagbarnes/sandbox

Considering both the objective and subjective truth within AI it should be noted that the two are interconnected. Whilst AI is mathematical and based entirely in logic, some ethical decisions cannot be boiled down to one thing or another, or a 1 and a 0 [5 & 6]. In these cases, events where AI would need to make a decision that could be considered ethical, it cannot truly be objective. [7]

For example, in recruitment AI could be seen to make a more objective decision [8], as it would have none of the unconscious bias' that traditional interviewers or selection processes have [9]. However, it has been seen that AI can predict characteristics or illness before the interviewee even knows about them [8]. This knowledge can then affect the decision as to whether to hire or not, as pregnant women would be more likely to cost the company more money than a man, for example. Obviously these features can be accounted for and turned off or coded out but if a business was looking to optimise it's work force, AI could be used to justify have a skewed workforce in favour particularly of men and physically and mentally healthy people ie. that it is all cost saving - despite there being research to suggest that a diverse and balanced workforce can lead to more productivity [8].

Additionally, AI can have biases coded into it. The most notable example is facial recognition software recognising images of black women [10&11]. In the majority of cases black women are identified as men and there is a strong suggestion that this is down to unconscious bias from the majority of engineers and computer scientists being male and white [11].

As has been discussed, AI struggles to be truly objective when presented with problems that have ethical questions tied to them [12]. From an interdisciplinary point of view, when facing ethical issues with AI it is important to be methodical and mathematical to avoid unconscious bias but it is also important to consider the wider implications of having AI make certain decisions, particularly ones involving human lives, either literally through self-driving cars and the trolley problem [13] or through the social impact of the decision, like selection criteria for a job.