I’ve just finished a Master’s project investigating the
movement of pathogens from the faeces of cattle to the sea where people
swim. The pattern of and processes
involved are fairly poorly understood, especially the movement of pathogens from
stream to sea. It’s the first time I’ve
been involved in studying such a complex and incompletely understood system and
it’s been great. With so much to learn
and the need for both a better knowledge of the patterns of pathogen abundance
and the need to better understand the processes responsible for I’ve found
value in pursuing and range of questions and approaches.
A large part of my research work involved regular field work
to collect data for a long term data set and the analysis of this data set. I could be highly confident that this work
would produce valuable insights. With data
collection and analysis planned out in advance I still had some spare time in
which to speculate, producing and testing hypotheses. I had room for many small failures. With the large knowledge gaps, curiosity and the
room to fail came a healthy attitude towards failure. Being aware of my ignorance, and the
ignorance of the larger scientific community meant that finding processes to be
not significant genuinely felt like a valuable lesson learnt. The cost of learning these lessons was
low. Materials were extremely cheap, the
lab work for preliminary experiments took at most a few hours and I didn’t need to prove my hypotheses to be
correct, on top of the long term dataset work any findings were a bonus.
Having benefitted from this set up I’m a big fan of
differentiating one’s work into that which is extremely likely to produce
valuable knowledge (even if only moderate amounts) and that which is more
speculative. Google do something similar
with 20% time encouraging employees to spend 20% of their time working on
curiosity driven projects not related to the rest of their work (see Adapt by Tim Harford for an elaborated
discussion of this idea or his talk on the subject). Ideally research as a whole should be a space
for failure but funders want results, significant results are more likely to be
published and referenced than non-significant ones and unless we are extremely
self-confident we struggle to do research without wanting to prove our ideas
(and therefore ourselves) to be right.
One solution is to have so many ideas you can’t get attached to any of
them, another is to view producing hypotheses like making bets, good bets often
lose, to paraphrase Nicolas Nassim Taleb, a mistake is not something to be identified
with hindsight but identified at the time with the information available. In producing a hypothesis we place a bet
using the information available. The
greater the rewards the more speculative our hypotheses can afford to be and
the more speculative we are they more we have to expect to disprove our
hypothesis.
