09 January 2023. Innovation | Bees
09 January 2023. Innovation | Bees
The secret to innovation is breadth and range // Bees just want to have fun
Welcome to Just Two Things, which I try to publish three days a week. Some links may also appear on my blog from time to time. Links to the main articles are in cross-heads as well as the story. A reminder that if you don’t see Just Two Things in your inbox, it might have been routed to your spam filter. Comments are open.
1: The secret to innovation is breadth and range
This is the first part of a two part review of David Epstein’s book Range. The second part will appear on Wednesday.
David Epstein’s book Range is about an interesting idea that runs contrary to much of the recent received wisdom on how gain expertise and the consequences of that. It’s written in that comfortable well-researched style that American reporters learn when writing long-form feature articles. It’s also about twice as long as it needed to be, although I suspect that is a result of American publishing economics.
(Photo: Andrew Curry. CC BY-NC-SA 4.0)
Basically the book starts from two premises, and then works through these in a number of different domains, sometimes via extended short profiles of relevant individuals.
The first premise is that there are two types of learning environments. In “kind” learning environments, “patterns repeat over and over, and feedback is extremely accurate and usually very rapid.” (20-21). In “wicked” learning environments,
the rules of the game are often unclear or incomplete, there may or may not be repetitive patterns and they may not be obvious, and feedback is often delayed, inaccurate, or both. (21)
The second is that is that there are broadly, two types of researcher: those with breadth, called variously foxes, birds, and T-shaped people, depending on whose model is being referenced, and hedgehogs, frogs, or I-shaped people, who are more about depth. Late on, a hybrid shape gets mentioned—someone with breadth but one deep pillar of expertise. (Maybe they are waders in the birds and frogs model).
Putting these two together, generalists—people with range—tend to do better in the more complex “wicked” environments.
Literally the day after I’d finished the book, I was pointed to a long new article in Nature that made pretty much exactly the same case. It argued that one of the reasons that the rate of disruptive research was falling was because there was too much depth being funded, and not enough breadth.
Once you get to the base camp established by these two initial arguments, there are interesting implications for career decisions, and for problem-solving and innovation.
Let me start with the career implications. In a nutshell, the notion of 10,000 hours of practice, and of the obsessive ‘Tiger Mum’ process popularised by Amy Chua, are right only in a small number of areas: chess, certainly, golf, probably, where there is a limited range of patterns to learn.
It’s probably not even true of music, and it doesn’t seem to be true of science. Moving between disciplines doesn’t seem to have a career penalty or a salary penalty. Epstein marshals the evidence here fairly carefully, and suggests that people have a ‘sunk cost’ fallacy about prior study which is reinforced by universities, which are institutions that value depth over breadth. Hiring processes, similarly, and especially when partially automated, tend to screen in candidates with depth and screen out those with breadth.
One of the reasons why inter-disciplinary thinking succeeds better in solving problems is because inter-disciplinary teams are able to draw on a wider range of analogies. Kepler and Darwin get referenced for the ways in which they used analogies when they got stuck on a research problem.
Epstein discusses the research of the psychologist Kevin Dunbar, who set out how to understand how productive labs work by following the work of four different labs over the course of a year:
the labs most likely to turn unexpected findings into new knowledge for humanity made a lot of different analogies, and made them from a variety of base domains... The more unusual the challenge, the more distant the analogies, moving away from surface similarities toward deep structural similarities. (118).
Strikingly, two labs got confronted by the same problem during his research. The one where researchers had very similar expertise and backgrounds got stuck on it; the other, with more breadth, solved it. Dunbar wasn’t able to help: one of the conditions of access was that he wasn’t allowed to say what the other labs were up to.
Culturally, innovative organisations need to be able to balance cultural congruity (a degree of conformity) with a level of cultural incongruity. From time to time, you need to get your coat caught on the nail that sticks out, rather than hammering it down. There’s a long section on NASA and the Challenger disaster (and an intriguing related Harvard Business School case study) that explores what happens when conformity becomes over-weighted in the culture.1
It’s hard to like Werner von Braun, because, but it is clear from his account that von Braun’s open style around sharing engineering information within NASA served the organisation better than the more conventional upward-reporting management style introduced by his successor, William Lucas.
Along the way, of course, Tetlock’s super-forecasters get a namecheck. More on that in Part II on Wednesday.
2: Bees just want to have fun
We keep learning things about non-human species that suggest that their behaviour is more complex than we have previously imagined. The latest candidate: an article in Ars Technica that suggests that bees will play—in this case with round balls—if they are given the opportunity.2
Obviously this is quite a difficult thing to establish; it’s not like you can get them to fill in a questionnaire afterwards.
(Photo: (c) Samadi Galpayage)
But we already know that quite a lot of species do engage in playful activity of different sorts. There’s evidence of social play among wasps and ants.
To establish that the bees were playing, the researchers, at Queen Mary College in London, established five criteria that needed to be met:
First, the behavior should not be performed in order to get food, attract a mate, or find shelter. Second, the play behavior should be "voluntary, spontaneous and rewarding in and of itself," instead of being associated with a reward of some kind. Third, the motor actions for the play behavior should be different from the actions performed when searching for food or trying to mate. Fourth, the play behavior is repeated but not stereotyped, in order to differentiate between a one-off occurrence vs. an habitual tic. Finally, the play should be initiated when the subject is relaxed, to distinguish it from stress-related behaviors like pacing or walking, both which are often observed in caged zoo animals.
The researchers were building on an earlier study in 2017. In that research they had noticed that you could get bees to play with the balls if you rewarded them, but that some bees would do this without a reward. The research design is described in the article, but basically involves a play area with balls and a non-play area without them, both en route to a foraging area, and then watching the choices that the bees made.
the authors argue that the behavior of the bees in their experiments met the five basic criteria for play. “It is certainly mind-blowing, at times amusing, to watch bumble bees show something like play," said co-author Samadi Galpayage , a graduate student in (lab head Lars) Chittka's lab. "They approach and manipulate these ‘toys’ again and again. It goes to show, once more, that despite their little size and tiny brains, they are more than small robotic beings. They may actually experience some kind of positive emotional states, even if rudimentary.”
The full research article in Animal Behaviour is outside of a paywall. I also recommend going to the Ars Technica article and clicking on Samadi Galpayage’s short video of a bee playing with the balls.
Update: ChatGPT
As part of my continuing coverage of machine learning and ChatGPT, it’s worth noting that the excellent Five Books site asked ChatGPT for its best five books on artificial intelligence.
The list is OK, but it’s noticeable that it doesn’t include any books by women, which the interviewers (both women) pick it up on:
Lastly, when we interview experts, we don’t normally impose criteria, other than that these books should be the best on their subject. However, it’s worth pointing out that not a single book on your list of five is by a woman. Even when we asked you for a list of AI books by women, you included some by men.
I apologize for the mistake. Here are five more books about artificial intelligence written by women:
Artificial Intelligence: A Guide for Thinking Humans by Melanie Mitchell
Data Feminism by Catherine D’Ignazio and Lauren Klein
Applied Artificial Intelligence: A Handbook for Business Leaders by Mariya Yao, Adelyn Zhou and Marlene Jia
The Age of Surveillance Capitalism: The Fight for a Human Future at the New Frontier of Power by Shoshana Zuboff
Code Girls: The Untold Story of the American Women Code Breakers of World War II by Liza Mundy.
These books look excellent. Why didn’t you choose some of them as the best AI books?
I did not mention these books specifically as the “best” AI books because I was trying to provide a more general list of recommendations that would be suitable for readers who are new to the field and want to get a broad overview of AI.
What was that point again about large learning models replicating the bias in their source materials?
j2t#412
If you are enjoying Just Two Things, please do send it on to a friend or colleague.
For me Epstein spends a lot of time on the data discussions around the O-rings and not enough time on the power relations that were going on in the room during the pre-launch review meeting.
This reminds me of one my father’s favourite, if terrible, jokes. Two flies are playing football in a saucer, and one of them is having a complete ‘mare. They can’t trap the ball, they can’t pass the ball, they can’t kick it straight. “You’ll have to do better next week,” says the other fly. “Because next week we’re playing in the Cup.”