avatarKoen Smets

Summarize

Featured image: via DALL·E

Dramatic dilemmas and AI

Can artificial intelligence make better decisions than we would ourselves? For that, it might be a good idea to look at the hardest decisions we face. And where better to look than in gripping drama?

We make a lot of decisions every day. The sometimes-cited figure of 35,000 seems just a little exaggerated (assuming a typical sixteen waking hours, it corresponds with a decision every two seconds), though. Even so, a lot of these decisions are simple — which sock to put on first, whether or not to look in the direction of an unexpected sound, or should we scratch that sudden itch (yes, in the privacy of our own home; no, during a job interview). Such decisions are of modest impact, and require little cognitive effort.

When trade-offs don’t give the answer

But some decisions are much harder — and much more interesting. These are the decisions that define good drama — whether it is an ancient Greek tragedy by Sophocles, a Shakespeare comedy, or a modern TV thriller. What powers the best narratives, even more than the dialogue, is the choices faced by the characters that populate them, and that they must make. The finest playwrights were (and are) invariably excellent observers of human decision making, who skilfully select tough challenges that resonate with the audience, and use them to bring the various roles to life. How do they face up to these challenges? How do they struggle and prevaricate, or alternatively how they act on impulse? How do they deal with the consequences? This is what expresses their humanity and what evokes their emotions in us, what makes us empathize with a character, or fuels our antipathy for them. What is quite remarkable is that we do not need to have had direct identical experience of the characters’ predicaments, to get a sense of the emotional turmoil that they undergo. If writer and actor have done their job well, we are — almost literally — moved by what we see and hear.

Good drama makes an emotional connection (image via DALL·E)

I have been enjoying Afterglow (Etterglød), a Norwegian drama series that follows a bunch of thirty-somethings. It is rich in exactly the kind of situations that produce this profound sense of recognition, even if we have never been in the same position. One of the many story lines is centred around the family of Ester, a hospice carer with a vivacious temperament, her husband Arild, a slightly more reserved astrophysicist, and their three children, Lars, Kira and Maya. The series starts on Ester’s 40th birthday who, true to her bubbly personality, has organized a large fancy dress garden party. But just as the first guests arrive, she receives a call with the news that her most recent smear test indicates she has cervical cancer.

A big shock, but she starts her treatment with great courage, trusting her positive mindset to guide her through the months of chemotherapy ahead. As the chemo sessions follow each other, however, Ester is finding it ever harder to cope, and little is left of her erstwhile spirited disposition. Before the diagnosis, the family had been planning to accompany Arild on a business trip to Japan, but this is now out of the question. As its timing coincides with that of Ester’s recovery after her final chemo treatment, both of them face a tough decision to make. Should he pull out and take time off to stay with his wife, or attend the prestigious conference? Should she ask him not to go, or instead insist that he goes on the trip as planned?

Such a decision is not easy, because it does not lend itself to conventional trading off of costs and benefits. It involves concepts like obligations, entitlement, commitment, expectations and duty, which do not let themselves be captured in cool calculation. Of necessity, we rely on our emotions to work out the best choice in situations like this — and so it is precisely in such scenes that good drama shines. They are both torn, but eventually Ester says that she wants her husband to go to Japan.

A machine’s deficiency

The day before Arild’s trip, we see Ester boarding the hospital’s minibus up for her final day in the chemo ward. As he is busy packing, Arild could not take her, but he promises he’ll pick her up in the afternoon. However, just as the nurse is about to connect the chemical cocktail to the catheter, Ester decides she is not going to have the final session. Many people — including her husband as she arrives home by taxi — would consider this an incomprehensible, if not irrational, decision. How on earth can you give up, so close to the finish line? Why not make this one final sacrifice in return for maximizing the chances of survival? Challenged by Arild, Ester expresses how she feels, and why: “I can’t take any more. If only you knew how awful it is…”. He snaps back, “Don’t you think I can see that? Do you think it is all about you?”

Arild, despite his unquestionable involvement, is nonetheless an outsider to this decision. And their argument shows how, as outsiders, we are easily led down the path of trade-off analysis, consider upsides and downsides and nett them off, usually in an ‘obvious’ way, with either costs vastly outweighing benefits, or vice versa. He can see that she has it tough, but in his eyes that is clearly eclipsed by the benefit of a completed treatment and the resulting improved chances of successful remission. Yet, that is not how Ester experiences the situation. To her, “I can’t take any more” means that an absolute limit has been reached. There is nothing more to give, no more trade-offs to be made. Or are there? Decisions always have stakeholders, who are not participating in the decision, but who are bearing its consequences. Did she properly take into account the loss her death would mean to her husband and their three children? Should she struggle on?

Go on, or quit? (photo: screenshot BBC)

When we are looking at such decisions, it is interesting to ponder about what use artificial intelligence such as ChatGPT would be in this context. My intuition is: not much, for two reasons. The first is that such decisions go way beyond weighing up costs and benefits that can be quantified or estimated, because they must take into account the obligations and entitlements mentioned before. This creates a complex constellation of, on the one hand, situational factors, and on the other the decision-maker’s own personality and, in particular, their values, which interact with each other in many ways. The same person will, under different circumstances, very likely make a different decision. At the same time, different people facing identical situations will also make different decisions. Exactly what we, in a given situation, take into account to decide, for example, whether to stop or continue chemotherapy is unknown, and arguably unknowable, to a machine, and so it cannot really make the call on our behalf.

But just imagine this insight could somehow be made available to ChatGPT, in the same way that we can — to some extent at least — reveal our values to other people by speaking to them, and having them observe and quiz us, so they can figure out what is important for us and how this would lead us to our conclusion. They would do that, just like we would, through a combination of reasoning and emotion — with emotion playing the lead role. Ultimately, we are unable to show our workings for such decisions. When we evaluate the moral rules, no matter how conflicted we are, we do not come to a conclusion by calculation and logic. We feel that a particular choice is the right one.

This is the second, and more fundamental reason. In the end, intelligence alone is not enough to resolve such dilemmas. Perhaps, one day, machines will be capable of being moved by compelling, gripping drama, and of feeling the grinding pain of impossible, dramatic dilemmas. Until then, we should be very careful if we intend to trust them with making decisions on our behalf.

Originally published at http://koenfucius.wordpress.com on June 30, 2023.

Thank you for reading this article — I hope you enjoyed it. Feel free to share it (the ‘share’ button below or at the top has direct options for Twitter, Facebook and LinkedIn) or simply copy and paste this link. See all my other articles featuring observations of odd human behaviour (I have been publishing one every Friday since 2016) here. Thanks!

Behavioral Economics
Decision Making
Psychology
Morality
Recommended from ReadMedium