29th of August 2019
Parkinson’s and the Flatland Press
A tale about escaping our cognitive limitations
Our methods of inquiry as scientists should not be determined by the ease with which we can communicate our work. Rather, our methods of inquiry should strive to directly tackle the immense complexity of our discipline while making it easier to collaborate, share, and develop our cumulative understanding | The Flatland Fallacy: Moving Beyond Low–Dimensional Thinking (Jolly & Cheng, 2018)
Few books that I have read impressed me more than Edwin Abbott’s book ‘Flatland’, written in 1884. Flatlanders live in a country with only two dimensions. One day a ball from the third-dimension falls through Flatland. For a resident of Flatland this looks something like this: First, a dot appears which transforms into an ever-increasing circle, after which the circle becomes smaller and smaller until it disappears.
One of the inhabitants of Flatland who goes by the name ‘A. Square’ argues that these appearing and disappearing circles must mean that the real world is complex and multi-dimensional, but is imprisoned because the rest of the inhabitants are unable to follow this line of reasoning.
The message that this story entails is, of course, that we – three-dimensional beings – are also limited in our cognitive abilities to understand complex reality. And even though we can understand that we don’t understand complexity – especially if we look into the universe on a beautiful summer evening, draw lines between the stars and ask ourselves whether alien life exists – in practice we prefer linear, easy to tell stories that make us believe that we do understand.
The same holds true for scientific research into Parkinson’s disease. Almost every article I’ve read says: Parkinson’s is a complex disease. Yet many of those same scientific articles then offer brave attempts to pretend Parkinson’s isn’t a complex disease after all and reduce its complexity to – preferably – binary questions. These questions are then answered with linear regression which plots a linear relationship between the variables studied. The type of graph I often encounter in the literature looks something like this:
The inescapable average
Yesterday, for example, I read the article with the binary question: ‘Parkinson’s Disease in Women and Men: What’s the Difference?’ (Cerri, e.a. 2019). The authors had – literally – drawn an image of the statistically average male and female Parkinson’s patient, based on information from graphs such as the above.
Seriously. How can patients defend themselves against this kind of imagery? In flatland charts, the individual – for whom I believe scientists ultimately work – is always lost. And that means that people like David Sangster, for example, have no other choice than to continue to fight for those who really do not fit into that average image. In this case the young Parkinson’s patients. But all of the efforts which focus on a new binary category – in this case the young versus the old – wouldn’t be necessary if we were to treat everyone as an individual with specific needs. Whether that individual is male, female, young, old, small, large, fat, thin, black, white, highly educated, without education or everything in between the binaries.
Because researchers themselves, of course, also know that the mere fact that you are a man or woman isn’t all-decisive (except for answering the question which of the two is pregnant), flatland research always ends with the conclusion that more of this type of research is necessary to be sure that the conclusions are correct.
But do we really need such follow-up investigations? Will it get us where we want to be? Because even though our brains usually tell us something else: Parkinson’s disease does not live in Flatland.
I would like to quote the words of Allen Newel who foresaw – in 1973! – that ‘we will not get there’ by continually doing research in which only a number of variables are examined. In this case Newel is talking about psychological research but in my opinion the same holds true for this type of Parkinson’s research.
The proper tactic is to frame a general question, hopefully binary, that can be attacked experimentally. Having settled that bits-worth, one can proceed to the next. The policy appears optimal – one never risks much, there is feedback from nature at every step, and progress is inevitable. Unfortunately, the questions never seem to be really answered, the strategy does not seem to work…… I am worried that our efforts, even the excellent ones I see occurring here, will not add up….. Maybe we should cooperate in working on larger experimental wholes than we now do | Newell, 1973
Larger experimental wholes… How visionary for someone from the pre-digital era!
In search for a possibility to escape, I found The Flatland Fallacy: Moving Beyond Low–Dimensional Thinking (Jolly & Chang, 2018). A feast of recognition! The authors indicate that we can move beyond Flatland, but only if we are open to acknowledging our own cognitive limitations.
The way out looks something like this:
- Be honest
Reflect on current research practice and be honest. Are we really going to get where we want to be like this? If so, where are we going to arrive? Be open to the fact that things could be improved. Even if you do not yet know how you should initiate such improvements and if there are practical objections and established research cultures in the way.
- Adjust to new realities
- Adapt research, research methods and ways to analyze data to state-of-the-art possibilities. E.g. use computer models and machine learning. Computers can reason in multiple dimensions whereas – much to my regret – we are lousy at it.
Models serve as tools to both reason and communicate about high–dimensional spaces. Models allow researchers to consider what dimensions of a problem are most relevant and predict outcomes based on complex sets of interactions. Moreover, models can be shared between researchers, permitting the collective development of a cumulative science whereby weak or redundant theories are pruned and robust, predictive theories are retained | The Flatland Fallacy: Moving Beyond Low–Dimensional Thinking (Jolly & Cheng, 2018)
- Collect as much, interoperable, data as possible and make it available to others. Only in this way can we fully exploit the potential of machine learning (Wise et al., 2019).
- Do not only collect large amounts of data but also work on enlarging data diversity. Collect data that you do not know or even expect to contain meaningful information. Then let the data speak for themselves.
- Train the future generation of researchers to make science “expandable”. Expand their training with skills such as the generation of predictive models, machine learning, data visualization and open science.
- Learn to value research differently
Nurture a culture in which the development, testing, sharing and further development of models is key. Have research output peer reviewed on “expandability.” Can other researchers/computers easily carry out the follow-up research recommended in the final paragraph of a scientific article?
The truth doesn’t care about our proof | Unknown
Jolly, E., Chang, L.J. (2018). The Flatland Fallacy: Moving Beyond Low–Dimensional Thinking. https://doi.org/10.1111/tops.12404
Newel, A. (1973). You can ’t play 20 questions with nature and win: projective comments on the papers of this symposium. W. G. Chase (ed.) Visual Information Processing, New York: Academic Press. https://pdfs.semanticscholar.org/85a0/96908670cd83cacfdede9e11f2df2dc41c9b.pdf
Wise, et al. (2019). Implementation and relevance of FAIR data principles in biopharmaceutical R&D. Drug Discovery Today, 24, pp. 933-938. https://doi.org/10.1016/j.drudis.2019.01.008