PhiloComp.net

Computers and Ethics

In the early years of this website (since 2006), relatively few people saw any special connection between Computing and Ethics. Discussions and outreach talks regarding the new Oxford degree in Computer Science & Philosophy highlighted Ethics as an obvious and important area of overlap (alongside Logic and AI), but there was little sense of any widespread moral concern. All this began to change as the Internet grew in importance within the lives of people, communities, companies, and nations, and shifted dramatically once the huge potential of machine learning applied to "big data" became clear. By 2020, indeed, it would seem odd not to put a major focus on Ethics in a website devoted to Computing and Philosophy, and accordingly this Ethics section now replaces what was just a single (long) page. Some of the pages here are currently rather short, but these will grow before long, as the entire section is under active development. For well-informed and substantial discussions of many of the issues, see the Stanford Encyclopedia links on Computer Ethics. It is striking that during the first decade of that monumental reference work (until 2008), only two of the eight listed articles had been written. Moreover five of the eight date from 2012 or later, thus again witnessing the spectacular flourishing of interest in this previously neglected field.

Oxford's New Institute

Oxford University has recently started a major "Ethics in AI" initiative, and is in the process of founding an Institute for Ethics in AI, whose members will be providing input to this website. Links to various existing materials can be found from Oxford Seminars on Ethics in AI, and many of these links have been incorporated within the pages that follow. The first such seminar began with two talks of more general relevance, one by Professor Sir Nigel Shadbolt, who has been leading this new initiative within Oxford, and one by Dr Carissa Véliz, who has generously contributed much of the text on specific topics here:

Mapping the Area

Organising the topics within this interdisciplinary area is far from straightforward, partly because of the complexity of the relationship between ourselves and "artificial intelligences". They are created by us and used by us, yet can also – in at least some minimal sense – act as "agents" themselves. Whether they are capable of genuine agency is controversial, but the very fact that they can mimic so much of the planning and reasoning that seems crucial to our own decision-making itself provides a further aspect of moral interest, in the reflection it prompts on our own moral capacities. Quite apart from these theoretical concerns, "intelligent" computer systems increasingly impact on our lives, both individually and collectively, in a wide variety of contexts (e.g. administrative, commercial, employment, entertainment, financial, legal, media, medical, military, social), and often in ways that go well beyond what was anticipated by their creators. All these things together make this one of the most complex and entangled topic areas, in both theoretical and applied ethics.

Bearing in mind that many topics may extend over more than one of the following categories, here is the structure around which the discussion in this section has been organised:

  1. AI as our creation
    Ethical issues in the research and creation of AI, including professional conduct, risks, intellectual property etc. Professional responsibilities will often depend, of course, on moral or other assessments that fall under later categories (e.g. if "superintelligence" is a serious existential risk, then researchers plausibly have a duty to avoid bringing it about).
     
  2. AI as an "agent"
    Concerned with the moral behaviour of robots as "agents" (e.g. self-driving cars, home helps, automated medical or military equipment), analogous to standard moral issues that arise in respect of human behaviour, but requiring precise codification in advance (quite unlike the mostly vague ethical precepts that we tend to rely on).
     
  3. AI as a decision-maker
    Most current AI systems are not robotic, but software within information systems that either make or advise on decisions, for example financial (e.g. whether to grant a loan) or legal (e.g. whether to allow bail). Such systems, especially when based on machine learning, can raise serious concerns about unfairness (e.g. owing to bias and opacity).
     
  4. Personal impact of computer systems
    Computer systems can often impact significantly on individuals who either use them or are affected by them. Prominent examples would be enhancement or impairment of one's autonomy through using an AI advice system; and loss of one's privacy (or other effects on social relationships) from security breaches or machine learning on "big data".
     
  5. Societal impact of computer systems
    Widespread use of computer systems can also have a major impact on society as a whole, for example by changing employment patterns, or by significantly affecting the distribution of wealth, information, and power. We have also seen how they can impact on the media, both public (e.g. "fake news") and private (e.g. targeted propaganda).
     
  6. Computer systems in medicine
    The field of medicine is something of a special case, owing to its importance throughout the various stages of our lives, its organisation (as involving a huge structure of cross-society investment, professional training and practice, research, commercial involvement, and public administration), and the wide range of ethical issues that arise within it.
     
  7. AI regulation and law
    There is a complex relationship between ethics and regulation: we don't consider all good actions to be compulsory (e.g. giving to charity), nor all bad actions to be forbidden (e.g. being rude). But where to draw these lines is especially difficult in the field of AI, given its novelty, speed of development, and complex variety of players and impacts.
     
  8. Fundamental philosophy of mind
    We have already noted the question of whether AI systems can exhibit genuine "agency". Other fundamental questions include whether (or how far) a computer system can be free, autonomous, conscious, have "personal identity", or be worthy of moral concern. These discussions can also shed important light on relevant aspects of our own nature.
     
  9. Meta-ethics and the nature of morality
    Computer modelling has played a major role in game-theoretic accounts of the evolution of behaviour, which are increasingly influential in contemporary attempts to provide a naturalistic foundation of morality. Thus even if we reject the idea of computers as having either moral agency or interests, they can provide significant input to moral philosophy.