Tuesday 30 June 2020

Which is the winning strategy - copying or innovating?

The question of whether innovation or copying is a more effective social learning strategy was tested a few years ago in an online tournament. The result may surprise you. 

Day 187
Originally uploaded by pasukaru76
Over a decade ago, the European Commission sponsored a research project called Cultaptation, to study the success of social learning strategies. This involved researchers from four universities in three European countries: Sweden, Italy and Scotland. It was coordinated by Stockholm University, where researchers from several departments formed a multidisciplinary “Centre for Cultural Evolution”.

The project explored social learning strategies by setting up an online tournament, with a generous proize,  to test out which strategy would be successful. The competitors were software "agents" who developed a repertoire of actions to survive in a changing landscape, and the contest looked for agents whose chosen learning strategy gave them the greatest survival rate. The learning strategies were based on variables such as the choice between

  • exploiting (continuing with actions in the current repertoire),
  • observing (learning from the behaviour of others), 
  • and innovating (trying something new).

The competition had a clear and outright winner, as shown in the plot below and described in the competition results paper.

This winner was DiscountMachine, which used the following strategy:

"Our creature does three major things: First it estimates/calculates, what we believe to be all the pertinent parameters of the simulation as well as a few other quantities that we believe to be useful.... Second it uses some of these parameters to estimate the expected payoff for performing each action in its repertoire. Once it has a best exploit chosen from its repertoire it compares the value of Exploiting to the value of Observing...... Lastly a machine learned function, takes into account N_observe and the estimates on the reliability of observing and P_c to adjust the value of Observing accordingly. Our creature then chooses whichever action has the higher perceived value, Observing or Exploiting. As a side note our creature only Innovates when it has an empty repertoire and observe doesn't work, which typically is only on the first turn of a simulation".

In other words, Discount Machine analysed carefully, all the time, whether to Exploit  (using its current best approach) or to Observe and learn from others, but it almost never Innovated.

An Article on the competition in New Scientist Magazine made the following observations:
"It seems a successful strategy rests primarily on the amount of social learning involved, with the most successful agents spending almost all their learning time observing rather than innovating"

"Avoiding spending too much time learning either socially or individually was
just as important. Between a tenth and a fifth of their life seemed to be the
optimal range. If they did more learning than that it seemed that life was just
passing them by."

"Successful strategies were also good at spacing out learning throughout the agents' lives. The winning strategy, Discount Machine, stood out because it did just this. It seems packing all your learning into the early part of your life is not a great idea - we need to keep updating our knowledge as we go along".

"You don't need any clever copying rules. You can just copy anyone at random. Other individuals are doing the filtering for you. They will have tried out a number of behaviours and they will tend to perform the ones which are reaping the highest rewards."

"To become the winner of the tournament you .... have to weigh up the relative costs and benefits of sticking with the behaviour that you have, versus inventing a new behaviour, versus copying others. That requires assessing how quickly the environment is changing, as this gives you an idea of how quickly information will become outdated".

"In variable environments (the winner) placed a higher value on more recently
acquired information and discounted older information more readily".

"Another attribute of the most successful strategies is that they are parasitic. This is the essence of social learning - somebody has to do the hard graft to find out how to do things before other people can copy them, so it only pays to learn socially when there are some innovators around. Indeed, in contests where (the winning) agents were able to invade the entire population, they actually ended up with a lower average pay-off than they did in contests where the conditions allowed some agents with more innovative strategies to survive, so providing new behaviours to copy".

Implications for Knowledge Management

For the knowledge manager, this is really useful and interesting experimental input. Given the emphasis on innovation you often see, it is good to be reminded of the value of copying as an effective competitive strategy. I remember one senior engineering manager telling me that his principle was "No Versions 1.0". He never wanted to be the guinea pig for new technology - innovation is risky, and sometimes the best strategy is to let others take the risks, and copy the survivors.

The need to adjust strategy depending on the variability of the environment is also interesting, suggesting that the variability of the knowledge context can affect the strategy needed to deal with it.

The ideal learning percentage is also an interesting statistic, also the fact that learning seems to be something that needs to happen constantly, rather than just at the start of your career.

The final learning point for me from all of this is that if we are looking at internal practices rather than competitive practices, then an organisation needs a knowledge management strategy that is strong on internal copying, leavened with a proportion of innovation.  If a team in an organisation meets a challenge, then (if they follow the DiscountMachine approach) they would look at their existing practices and compare them with practices elsewhere in the organisation.

  1. If their existing practices are as good or better, they should use the "Exploit" approach, and follow these existing practices. 
  2. If another team has a better practice, they should use the "Observe" approach, and copy the other team's practice. 
  3. Only when there is no existing knowledge that will sufficiently meet the challenge, should they use an Innovate approach.

This was the winning strategy for DiscountMachine, and it may well be the winning strategy for teams within an organisation.

No comments:

Blog Archive