There is a lot of science in coffee (and a lot of coffee in science). And there are also many scientists who are keen coffee drinkers and vice versa. But is there more in common between these two fields than even this? Could a shared problem be hiding a different (shared) problem?
One issue for coffee drinkers is reliability and reproducibility. How can we ensure that we get a good cup each time we visit a café or brew our own? In a similar way, how do we ensure that our experimental results in science are correct and reproducible? It is a fairly fundamental tenet of science that an experiment should be able to be reproduced in another lab with similar equipment. The suggestion is that this is not always happening, we have a ‘reliability’ problem in science (and sometimes in coffee).
A possible solution, in both fields, is some form of automation. In the world of coffee this is quite obviously just by making coffee via a machine. There have been several attempts to make reproducibly good coffee using an automated pour over machine. In the world of experimental science, it is not quite so clearly an automation process but is described instead as “data sharing”. The results of all experiments, or at least those that are published, should be shared (uploaded with the published paper) so that others can examine the data in more detail and form their own conclusions†.
For both coffee and science, it is suggested that this opening up of our process so that it is more transparent or reliable, will increase the reproducibility of good results and, crucially, mean that we get those results faster. We will get reproducibly good coffee without having to queue so long in the morning; we will make discoveries more quickly and have faster progress in science.
It seems that the problem that we thought we had, reproducibility, is perhaps not the one that we are actually aiming to solve. The problem we seem to be interested in is ensuring faster progress.
But an emphasis on faster progress can undermine our initial ideal of a reproducibly good result. For scientific research the emphasis on getting results quickly (at least within the time frame of the science funding cycle) has led to predictable problems. There are cases that I know about where results that were contrary to those that were wanted were suppressed. Not permanently. No, that would be demonstrably scientific fraud. No, suppressed just long enough for the first ‘ground breaking’ paper to be published. Then, after a suitable length of time, the second paper showing the problems with the first can follow up. Those involved get two papers (at least), and rapid progress is shown to be happening in the field. Would data transparency help here? Clearly not, because the initial set of data would still be suppressed until it was wanted those few months later.
What we need is a change in the structure of how science is done. We need to value scientific integrity and so trust that other scientists do too. We know that the current situation whereby promotions, funding etc are determined by the number of papers in ‘good’ journals, can act to undermine scientific integrity. This needs to change if the reliability issue is to be addressed. In the type of case described above, there would be no consequences for the people who kept their names on the paper(s) published. The only consequences would be if anyone refused to have their name on the paper as they knew it was misleading. And even then, the consequences would be to that person/those people in terms of their CV, and publication list, not those who published the paper and of course, shared the data.
For the coffee, there are already discussions about whether the increase in throughput offered by automated coffee brewing techniques really contributes to the coffee experience that the cafés are trying to encourage. Can we really expect someone to slow down, take in the aroma, the mouthfeel, the taste and flavour if we rush the cup through to them on a production line? Isn’t part of the enjoyment of something to have to wait for it (hence lent before Easter; fasting before a feast)?
It is not that automation necessarily is bad. We can get a genuinely all round good coffee in a café that utilises a machine based pour over (perhaps). We can also get genuinely reproducible data in a situation where data is routinely shared. It’s just that data sharing does not solve the reproducibility problem, nor does automation give us continually good coffee. What makes the difference is a café that cares about the product that they are serving; scientists that care about the integrity of the research that they are doing. Automation processes give us faster results, they do not, automatically, give us better science (coffee).
Moreover, our desire for faster progress obscures questions that we should be asking if we slowed down a little. What is a good coffee experience? Who (if anyone) should own the scientific data shared? Is our desire for good coffee, quickly and (relatively) cheaply obtained, an aspect of that consumerism that is damaging for the planet’s ecological health? How much do we need to trust each other (and take responsibility for our own integrity) for our society, including our scientific society, to function? Is faster progress in and of itself, a “good” to aim for?
And perhaps, there is a final, more fundamental question. Have we become so accustomed to seeing ourselves and our work as merely a cog in a machine that we have become inured to the dehumanisation of society which seems to us almost natural and itself progress? Is this what we want for society?
The process of making a good cup of coffee indeed shares many things with the process of doing science. Perhaps this should not be surprising, both are practises embedded in our society. Certainly our view of the society that we live in can be informed by slowing down with our coffee as we enjoy a little science (or should that be the other way round)?
†It is not quite an automation process in the sense that the data is taken by a machine and then uploaded. However, it is still a dehumanisation process. At the root of the concept is the idea that the human experimenter can be taken out of the process. I would be happy to expand on this in the comments but for the sake of readability haven’t done so here.