Watching Steen-Johnsen on “Concern and enthusiasm for AI across the globe: the role of trust”

I recently watched a really interesting seminar by Kari Steen-Johnsen (UiO), who is working on the impact of digitalization on democracy. She gave an AI Lund online seminar on “Concern and Enthusiasm for AI Across the Globe. The Role of Trust” (recording here).

The main question was how AI is received in different context across the globe. In general, there are very different perceptions in different contexts, AI is seen (and marketed) as the great equalizer by some, or as a threat to the world in terms of misinformation, job security, threat to citizens’ rights etc by others. But how AI systems are perceived depends, in their model, on citizen’s ability, motivation, cognitive bias, and AI trust. And how people perceive AI influences how they work with it, which can then influence their political knowledge (informed, uninformed, misinformed) and beliefs and attitudes (updated or rejected). This can then lead to democratic consequences, like political polarization.

So how does that depend on the context people are in? To answer that, Steen-Johnsen et al. looked into survey data for trust in AI from six countries (Brazil, Denmark, Japan, Netherlands, US, South Africa). But trust in AI is not an absolute measure; how important trust in AI is depends on how strong people trust in others at the same time, for example in institutions, especially when there is a conflict between the two. Steen-Johnsen et al. postulate that trust is a coping mechanism when confronted with rapidly changing world and risks and uncertainties, and assume positive association between trust and enthusiasm towards AI (which I find weird — their scale goes from concerned to enthusiastic, but to me those are not opposites? I am very concerned about intentions of developers/funders and missing control mechanisms and data privacy and citizen’s rights etc, at the same time I am enthusiastic about early detection of breast cancer and about new developments in technology. So I don’t even know what I would have responded, so I find it very difficult to interpret what others might have meant). Anyway, does trust towards AI depend on how trusting a person is in general, or is it really specific to circumstances?

Steen-Johnsen et al. find that on average, people are more concerned than enthusiastic, and the most enthusiastic in their daily life and most concerned for democracy, with concerns for economy and journalism in between. Looking at the distributions of trust in the different countries to see if there is polarization in the sample shows that all distributions peak somewhere in the middle, but Denmark, the Netherlands, and the US also have peaks at far end of concern. So far, so good: Context matters!

Next, they looked into trust in the information landscape (i.e. traditional media, Wiki, Social Media). The more people trust the information landscape and have political trust, the more they state they are enthusiastic about AI. This is most strongly for journalism, and least strongly for daily life. Interestingly, the two countries that are closest together geographically and probably also culturally, Denmark and Netherlands, are the countries on either end of the spectrum here. The relationship between trusting the media landscape and stronger trust in AI is stronger in the Netherlands, because “a rising tide raises all boats“, so in a high-trust country like Denmark that effect isn’t as visible (Or that was at least how I understood the explanation).

More results are that trust in ordinary people (“social trust”) bolsters against fear in other studies, but not here against concern about AI. Trust in science does not seem to play a role (except in Japan), and trust in legal institutions also does not play a role on how much or little people trust AI.

So far, so good!

I think the most important thing to note for my context is that all the western countries have similar perception of AI, but not in the trust relationships regarding AI. Context matters! And maybe, if we think about AI and trust, and relationships between teachers and students in the presence of AI, we need to pay more attention to where, for example, studies were done; on what background of trust in general in that context, but also trust in institutions (like universities, or institutions more generally). And when we think about trust-building behaviors that teachers can engage in, we need to also keep in mind that how willing or even able students are to extend trust to a teacher will also depend on their prior experiences with other teachers, maybe adults or people in positions of power in general, or institutions etc.. We haven’t looked at that at all and my point is mostly that it’s good to remember that it’s always more complicated :-D

Featured image and pics below from a couple of days ago…

Very windy day, very little water!

And dramatic skies…