These points are extracted from a combination of my notes and my memory, both of which are at best only noisy indicators of the true content. I imagine a large amount of editorialising is present. If anyone feels their views are misrepresented I’d be happy to include their comments alongside my orthogonal interpretation, so please do correct me!
Planning under cognitive limitations
Mehdi Keramati, who attended the previous summer school as a student and returned this year as a lecturer, gave a tour of different approaches to decomposing cognitively-intractable problems into solvable tasks. The familiar contrast between model-free (habit-based, inflexible) decision-making and model-based (cognitively resource-intensive, flexible) was invoked, with behavioural evidence to demonstrate that humans use hybrid strategies.
Alongside presenting the limitations of pure habit-based and model-based cognitive strategies, we were given a rundown of at least a handful of enticing hybrids: plan-until-habit which allows for easy assessment of familiar paths of the decision tree; and most enticingly bidirectional planning.
Bidirectional planning involves simultaneously working from position to goal and from goal to position, substantially narrowing the proportion of the decision tree which needs to be explored. The behavioural evidence offered to illustrate that humans use this kind of planning was especially neat. Participants were trained, through a combination of passive and active study, to memorise the social connectivity of a carefully-structured social network. In the test phase, participants were asked to nominate the three individuals through whom a message would have to pass to most efficiently connect two prompt individuals. Critically, participants were given the option of highlighting certain individuals in the choice array with different colours, and their use of this highlighting tool presented clear evidence of bidirectional planning: participants would first highlight individuals connected to the source individual in one colour, then highlight those connected to the destination individual in another colour, reducing the problem to finding an individual connecting two individuals highlighted in different colours.
In the subsequent discussion there was some debate over whether the distinction between habit-based and model-based strategies was more apparent than real, with habits being regarded by some as potentially explicable as chunked model-based solutions, although no clear consensus or detailed argument sticks in my memory.
What’s so great about working with stupid and useless people?
Bahador Bahrami gave a lighthearted but informative talk in which he explored group decision-making. He briefly presented a decade of work illustrating the benefits of working with others, provided those others are of similar decision-making ability to oneself. He also presented the equality bias findings wherein participants will overweight incompetent opinions and underweight competent opinions in the service of equality, even when they are provided with direct feedback indicating the difference in quality and offered financial rewards for maximising performance.
The dangers of equality bias, and the potential gains of working with others, together raised the question of how heterogenous the population is. Some empirical sampling from the hundreds of participants who have completed these joint decision-making tasks over the years, it was estimated that random parings are sufficiently similar to at least break even (have a joint sensitivity equal to their average sensitivity) around 90% of the time. An encouraging finding indeed (and answers the question posed in the title along the lines of “nothing, but they’re harder to find than you’d think”).1
The second part of the talk focused on group responsibility, and showed that people feel less responsibility for decisions when they make decisions as part of a group. This seemed in part to be a function of a resistance to regret, which consequently led to perseverance in the face of poor outcomes. Talking to Marwa El Zein after the talk, on whose work this latter part was based, it emerged that many properties of the decision-making task have profound implications for the subjective feeling of responsibility, including the outcome (people feel more responsible for good rather than bad outcomes).
Getting more from wine
Barry Smith finished off the week’s lectures with a philosophical analysis of wine tasting. Having provided a wine-tasting taster session just prior to the talk, the audience was particularly receptive to the content.2
In the line of analytic philosophy, some key distinctions were highlighted: wine tasting is about the wine, not the taster or the liking; and there’s room for both sense and nonsense in wine-tasting assessments. While never getting a clear and concise conception of what makes an expert wine taster, we nevertheless acquired some interesting suggestions: expert wine tasters are able to identify and focus on particular aspects of wine; they have an awareness of their own sensations and a vocabulary for expressing them; and their appreciation of wine is powerfully shaped by their prior expectations arising from an in-depth knowledge of grape-growing and wine-making processes.
We were offered some discussion of the tension between the democracy of taste, in which all opinions are equally valid, and the very notion of expertise evaporates, and the expert-as-guide position which asserts that there are objective3 facts about wine which are accessible to the wine expert by virtue of their skill and training. The presence of wine apps, which offer something of a middle-way by democratising expertise, indicate yet another approach, although the most prolific users of these apps have sufficient nuance and experience to be considered at least gifted amateurs in the wine-tasting game.
Despite some concepts being largely vacuous (‘minerality’ was singled out as essentially a meaningless general gesture of approval), and despite rather underwhelming performances in blind taste tests, I am somewhat more persuaded of the richness of the wine-tasting world than I was at the beginning of the day. Particularly interesting for me is the idea that novices and experts are performing qualitatively different tasks during a blind taste test: when a novice drinks wine they drink it in ignorance, while an expert drinks a wine with a wealth of expectations derived from their knowledge of the region, the vintage, the process, etc. To force an expert to taste as a novice might have a childish appeal, but one ought to be careful about drawing strong conclusions from such an exercise.
Overall I was left with the impression that wine expertise may be a little like Dennett’s response to Thomas Nagal’s essay “What is it like to be a bat?”: the novice may have a very idiosyncratic and inconsistent response to a wine, but they can follow the recommendations of an expert as a promise that, as their own expertise increases, their experience of the wine will begin to approximate that of the expert. By highlighting aspects of the experience to which the novice can attend, the expert offers them stepping stones on the journey to expertise.
I didn’t take photographs or notes of the many, many excellent posters that were presented during the week. I am grateful to those who took the time to explain their work to me and to discuss its implications and potential extensions. There was so much exciting and original work on display that I don’t think I inspected more than half the posters. Thanks to all the attendees who provided stimulating conversation, whether at your poster or by the sea.
It’s quite possibly true that the distribution of perceptual sensitivity in the population is more homogenous than the distribution of traits relevant for other kinds of decision-making tasks, particularly real-world tasks of key interest such as policy negotiation. ↩
I don’t drink, but it might be accepted that my natural enthusiasm for philosophical analyses rendered me likewise attentive. ↩
(Or at least inter-subjective) ↩