What Shapes Ethical AI Research?

A standardised way to teach ethics in Computing Science is missing. As the range of domains that AI effects continues to increase, we can’t let media be the sole decider on what constitutes good and bad. We need to think more foundationally.

When was the first time you came across artificial intelligence? For many it wasn’t watching a neural network classify cats or dogs, or a GAN generating artwork of whatever comes to mind. It was through media. Whether it be movies, books, artwork, AI has been explored in media for decades if not close to a century at this point. Even for those not actively working in the field, AI continues to fascinate us with ‘what-ifs’ and ‘what-nows’.

And for those working in the field, media plays a key role in shaping what problems are worked on, how they’re approached, and critically, influences ethical reasoning. As Dillon and Schaffer-Goddard explore in their paper from earlier this year, the reading choices of AI researchers often played a significant part in influencing their work. Interviewing 20 current AI researchers in the UK, the role of media in promoting ethical reasoning was widely noted by many of the participants.

Some drew inspiration from the likes of Dr Who where, in one particular episode, the Doctor must decide whether or not to destroy the Daleks - his mortal enemy. Does he destroy a creature that he knows will bring suffering and destruction or let them live as a timeline without the Daleks could lead to something worse? The parallels with current concerns in AI research are obvious. Several others saw media as a way to guide the kind of research they worked on. Most notably, multiple researchers cited frequently reading dystopian and speculative fiction where the potential negative outcomes of technology are often depicted. In more than one instance, researchers stated that this deterred them from pursuing work on facial recognition or other privacy-invasive technologies.

While it’s reassuring that the researchers interviewed generally saw media as a positive force in their work, for the good of the field and for Computing Science as a whole, we need to think more foundationally. We can’t let media be the dominant way that those in Computing Science come to determine ethics. As it currently stands, teaching ethics is largely left out of many Computing-related degrees; and when ethics is covered, it can be relegated to a one-off session or left as simply ‘further reading’. How ethics should be covered is also lacking standardisation meaning different Computing courses cover different views in different ways for different lengths of time. Such variance results in a simple question: What do we teach when we teach ethics? Indeed this is what Fiesler, Gerrett, and Beard open their paper from SIGCSE with.

Reviewing 115 different syllabi from various Universities, there emerges a surprising variance in how ethics is taught. The majority of classes focused on the law and policy aspect of ethics: what are the legal aspects of working in Computing Science? Others took to focusing on privacy and surveillance: the role of big data, social networks, and digital footprints. Thirdly, philosophy was a major focus: how should the work of Computing Scientists be applied in the context of a philosophical framework?

As the framing of ethics varied between courses, so too did intended goal. Almost three quarters of courses used the teaching of ethics as a way to encourage critical thinking of Computing Science work. Others sought to encourage students to be better equipped to notice when a system is behaving unethically. Another major area of focus was helping students better communicate how systems should operate and make the ethical concerns of such systems explicit.

Ethics should be an integral part of all Computing-related degrees. While it’s reassuring that some Universities have started to teach ethics in one way or another, there’s still much work to be done. Whether it be AI, Big Data, Cybersecurity, or a host of other subfields, the domains in which Computer Scientists work are having ever greater impact. It’s only right then that such work be ethically conscious.

Further Reading

  • What AI researchers read: the role of literature in artificial intelligence research
    • Authors: Sarah Dillon, Jennifer Schaffer-Goddard
    • DOI: 10.1080/03080188.2022.2079214
    • Open Access: Yes
    • Available here.
  • What Do We Teach When We Teach Tech Ethics?: A Syllabi Analysis
    • Authors: Casey Fiesler, Natalie Garrett, Nathan Beard
    • DOI: 10.1145/3328778.3366825
    • Open Access: No
    • Available here.

Published as CC BY-SA 4.0.