Where ethics and decision making overlap…

Where ethics and decision making overlap…

From the AiM Next Generation Technologies team

Let us consider two ways of making decisions. First, we gather all relevant information, maybe in several iterations, and using a consistent approach produce a ‘considered decision’. Second, we use less relevant and inconsistently applied questions, to reach a decision which ‘feels good’. Humans generally make decisions based on both methods. In addition, as well as using random methods, we do this randomly, so there is rarely consistency in the decision-making process. Exceptions exist – years of training and experience allow professionals to apply similar question sets and criteria to problems and come to consistent answers, which drive decision making, for example in medicine, engineering and cookery. However, even when humans use consistent decision-making processes, our capacity to apply all the data available is limited.

AI on the other hand, uses algorithms to make consistent decisions, based on models formed by learning from historic data; and the more data, the better the decisions. Additionally, AI does not need a rest, and will follow the same model, regardless of having done it a thousand times before. And note, the use of AI is not designed to replace a human, rather it complements them, and expands their understanding of the subject area, allowing better interpretation of the results.

So far so good, what could go wrong? Well a key component is the ‘learning’ phase, where the AI either via supervised or unsupervised learning, determines the rules it will use to make decisions. Could these materials be manipulated to drive predetermined ‘biased’ results? It’s possible if the model is set up or analysed in a way which leads to systematic bias; and herein lies the ethical dilemma. How do we ensure that the models we use are unbiased, and how much trust do we put in those making the decisions on what materials to use? After all, regardless of your morality, if your livelihood is dependent on a particular outcome, you are prone to being swayed.

So, we see that AI has great benefits, but we must be sure that how it’s implemented is based on the greater-good, rather than the benefit of the few. This begs the question, “how do we make the use of AI ethical?”. What do you think? Can it even be done? Tell us your views and join the debate on Twitter.