Roland Kupers wrote a magnificent book. A novel with a deep philosophical  message. October 15th 2025 was the official book launch. I was asked to give a short reflection on Zebras and Reductionism. Here it is (Dutch version below).
Zebras and Reductionism
 
To understand why a zebra cannot be domesticated — while a donkey or a horse usually can — Roland has his protagonist George Du Paon observe:
 
“… it is the impression of the whole we are after; we have a good idea of the parts from anatomic drawings that are readily available. The wildness is not expressed in the individual muscles, tendons, or skeleton.”
 
In other words: it’s about the impression of the whole. We are in good company here, because Goethe already put it beautifully in 1810: “Nature has neither core nor shell; she is everything at once.”
 
And more than 2,300 years earlier, Aristotle sighed that “the whole is greater than the sum of its parts.” In fact, we now understand the whole is not only greater, but fundamentally different from the sum of its parts (Philip W. Anderson, 2011).
 
And that is what I want to talk about.
 
In my field — Complex Adaptive Systems — this phenomenon is called irreducible emergence: the rise of new properties in a complex system that cannot possibly be explained by looking only at the components. Take such a system apart, and you lose precisely the property you were trying to understand. It’s like dissecting a cat to find out why it purrs: you’re left with a pile of parts — but the purr is gone (along with the cat, incidentally).
 
And yet we are all prone to this kind of simplification, which I believe is driven by a deep longing for Laplace’s causally deterministic universe. We are endlessly tempted to break complex problems into isolated puzzle pieces. Whether it’s climate change, social inequality, geopolitical tensions — or, for the enthusiast, all three at once — we keep trying to flatten complexity. As if knowing the parts well enough will automatically reveal the whole.
 
But that is a major misconception.
 
Take the human brain: some 80 billion neurons, each connected to tens of thousands of others. That number of connections is roughly on the order of the number of seconds since the Big Bang — 13.8 billion years ago. Together, these neurons produce our thoughts. But how much of a thought resides in a single neuron? A strange question. So how about in a hundred neurons? A million? A billion? Where is the phase transition, the regime shift? When does the question suddenly stop being strange? We simply don’t know. Thoughts are emergent properties of dynamic processes on that neural network. Just as our economy is an emergent property of everything we contribute together. Or just as the mesmerizing wave of a murmuration of starlings — meant to confuse predators — is an emergent property of interactions among many individual birds.
 
The physicist Murray Gell-Mann — Nobel laureate and father of the quark — once asked me, when we were celebrating his 80th birthday in Singapore: “How wet do you think a single water molecule is?”
 
It turns out to be quite the conversation starter at birthday parties…
 
It is therefore surprising — perhaps even a bit disappointing — that we keep falling back into reductionism. Even though the Dutch philosopher Benedictus de Spinoza warned us in 1677: “Every part of Nature agrees with the whole and is associated with all other parts.”
 
And yet René Descartes, in the very same period, casually claimed that all non-human animals are simply machines. A view echoed later, on an even grander scale, by Isaac Newton. Hardly intellectual lightweights, either.
 
In complex adaptive systems research, we try to get a grip on phenomena like emergence — by building new mathematical models and running computer simulations to see how information flows through a complex network, and at what point new characteristics begin to appear. This sometimes leads to bizarre discoveries — such as systems that grow stronger when partially destroyed, or the spontaneous collapse of causality in complex networks. But even here we must stay vigilant: in a successful simulation, the computer may understand what’s going on — long before we do.
 
Let me end with a warning.
 
This relapse into reductionism is accelerating — visible not only in the rise of populism, driven by oversimplified narratives on the internet, but also in the misguided faith we place in data and artificial intelligence. The belief that, if we only collect enough data, patterns will automatically emerge that reveal the underlying processes — as Hans Rosling’s Our World in Data tempts us to believe — is fundamentally flawed. This way of thinking is understandable, but also a major regression to the inductivism of Francis Bacon in the early 17th century, where observation and data were considered the sole path to knowledge. It leads, in my view, to an uncomfortable form of data fetishism.
 
Data — without underlying testable models — is dead. It is merely a record of what has already happened, and says anything about the underlying processes or the future only if the future is identical to the past. This unhealthy mix of lazy reductionism and flat inductivism seems to me one of the major causes of misunderstanding — and mismanagement — in the world today.
 
(And from experience, I can tell you — this is not a crowd-pleaser at birthday parties.)
 
Fortunately, there are thinkers like Roland Kupers, who invite us — with playfulness and at times outright hilarity — to reflect on these questions. Whether you read his book to learn how to tame zebras, or simply to tickle your brain: you will quickly notice that the wild side of reality is not so easily caged.
 
And yes — his book fully deserves the legendary inscription on the new library of the University of Amsterdam — my Alma Mater — written in 24 languages: “Lees maar, er staat niet wat er staat.” (“Just read — it does not say what it says.” — Martinus Nijhoff in Awater).
 
Falling in Love while Stuffing a Zebra is very much greater — and different — than the sum of the parts from which it is made.
-----

Zebra’s en reductionisme
 
Om grip te krijgen op de vraag waarom een zebra niet te temmen is – terwijl dat met een ezel of een paard doorgaans wél lukt – laat Roland zijn hoofdpersoon George Du Paon het volgende opmerken:
 
‘… it is the impression of the whole we are after, we have a good idea of the parts from anatomic drawings that are readily available. The wildness is not expressed in the individual muscles, tendons, or skeleton.’
 
Oftewel: het gaat om de indruk van het geheel. We zijn hier in goed gezelschap want Goethe zei daar al in 1810 iets prachtigs over: ‘Die Natur hat weder Kern noch Schale, sie ist alles auf einmal.’
 
En ruim 2300 jaar dáárvoor verzuchtte Aristoteles dat “het geheel méér is dan de som der delen.”  Sterker nog we begrijpen nu dat het geheel niet alleen méér, maar vooral ook ánders is dan de som der delen (Phillip W. Anderson 2011).
 
En dáár wil ik het even over hebben.
 
In mijn vakgebied -Complexe Adaptieve Systemen-, heet dit verschijnsel irreducibele emergentie: het ontstaan van nieuwe eigenschappen in een complex systeem welke je onmogelijk kunt verklaren door alleen maar naar de onderdelen te kijken. Trek je een complex systeem uit elkaar, dan verlies je precies die eigenschap waarnaar je op zoek bent. Het is alsof je een kat uit elkaar haalt om te begrijpen waarom ze spint: je houdt een hoop onderdelen over, maar de spin is weg (en de kat trouwens ook).
 
Toch hebben we allemaal de neiging tot dit soort vereenvoudiging, iets dat volgens mij gedreven wordt door een diep verlangen naar het causaal deterministische universum van Laplace. We laten ons telkens weer verleiden om complexe problemen te reduceren tot losse puzzelstukjes. Of het nu gaat om klimaatverandering, sociale ongelijkheid, geopolitieke uitdagingen – of, voor de echte liefhebber, alle drie tegelijk – telkens proberen we die complexiteit plat te slaan. In de veronderstelling dat als we de onderdelen maar goed genoeg kennen, het geheel vanzelf duidelijk wordt.
 
Maar dat is dus een groot misverstand.
 
Neem nu ons brein: zo’n 80 miljard neuronen, elk verbonden met tienduizenden andere. Dat aantal connecties komt al aardig in de buurt van het aantal seconden dat verstreken is sinds de oerknal, 13,8 miljard jaar geleden. Al die neuronen samen produceren onze gedachten. Maar hoeveel van een gedachte zit er eigenlijk in één neuron? Tja, een nogal vreemde vraag. Maar stel dan: hoeveel zit er in honderd neuronen? Of in een miljoen of een miljard? Waar zit die faseovergang, die regimeshift? Wanneer wordt het ineens géén vreemde vraag meer? We weten het simpelweg niet. Gedachten zijn emergente eigenschappen van dynamische processen op dat neuronennetwerk. Net zo goed als onze economie een emergente eigenschap is van wat wij allemaal samen bijdragen. Of zoals de macroscopische golfbeweging van een zwerm spreeuwen -die roofvogels in verwarring moet brengen- een emergente eigenschap is van de interactie tussen vele individuele vogels.
 
De natuurkundige Murray Gell-Mann – Nobelprijswinnaar elementaire deeltjes en ‘vader van de Quark’ – zei eens tegen mij toen we zijn 80ste verjaardag vierden in Singapore-: “Hoe nat denk je eigenlijk dat één molecuul water is?”
 
Dit is best wel een aardige gespreksstarter gebleken op verjaardagsfeestjes…
 
Het is dan ook verrassend, en misschien wel een beetje teleurstellend, dat we steeds weer terugvallen op dat reductionisme waar de Nederlander, Benedictus de Spinoza ons in 1677 nog voor waarschuwde toen hij schreef: “Every part of Nature agrees with the whole and is associated with all other parts.”
 
Maar René Descartes kwam in dezelfde periode doodleuk met de stelling dat alle niet-menselijke dieren simpelweg machines zijn. Een visie die we later in een nog groter verband ook bij Isaac Newton terugzien… En dat waren toch niet de minste denkers.
 
In het complexe adaptieve systemen onderzoek, proberen we grip te krijgen op dit soort fenomenen zoals emergentie. We doen dat door nieuwe wiskundige modellen waarbij we -via computer simulatie- kijken hoe informatie door een complex netwerk gaat en op welk moment er nieuwe karakteristieken ontstaan. Dat kan soms tot bizarre verschijnselen leiden, zoals de ontdekking dat een netwerk/systeem sterker kan worden door het juist gedeeltelijk te slopen of de ontdekking van het spontaan verlies van causaliteit in complexe netwerken. Maar ook hier moeten we alert zijn op het gevaar dat bij een succesvolle simulatie de computer misschien wel begrijpt hoe het werkt, maar wij nog lang niet…
 
Laat me afsluiten met een korte waarschuwing.
 
Die terugval naar reductionisme zien we steeds vaker. Niet alleen in de opkomst van het populisme -gedreven door al te makkelijke verklaringen op het internet, maar ook in de waarde die we hechten aan data en kunstmatige intelligentie. De gedachte dat we, als we maar genoeg data verzamelen, vanzelf patronen zullen ontdekken die ons de onderliggende processen onthullen – zoals Hans Rosling in Our World in Data ons wil doen geloven – is fundamenteel fout. Dit soort denken is begrijpelijk maar tegelijk een enorme terugval naar het inductivisme van Francis Bacon uit het begin van de 17de eeuw waarbij waarneming/data de enige bron naar kennis is, het leidt m.i. tot een ongemakkelijke vorm van datafetisjisme.
Data -zonder onderliggende toetsbare modellen- is een dood ding, het is een registratie van iets dat geweest is en dat alleen maar iets over de onderlinge processen of de toekomst zegt als de toekomst identiek is aan het verleden… Deze ongezonde combinatie van lui reductionisme enerzijds en plat inductivisme anderzijds lijkt mij een van de belangrijkste oorzaken van misverstanden en misstanden in de wereld…
 
(En uit ervaring weet ik dat dat nou net géén lekker onderwerp is op verjaardagsfeestjes.)
 
Gelukkig zijn er denkers zoals Roland Kupers, die ons op speelse en soms ronduit hilarische wijze laten nadenken over dit soort vragen. Of je zijn boek nu leest om zebra’s te temmen, of gewoon om je brein een beetje te kietelen: je zult merken dat de wilde kant van de werkelijkheid zich niet zo makkelijk in een hokje laat stoppen.
 
En zeker ook voor dit boek geldt wat er in 24 talen staat op de nieuwe bibliotheek van Universiteit van Amsterdam -mijn Alma Mater-: ‘Lees maar er staat niet wat er staat’ (Martinus Nijhof in Awater).
 
Falling in Love while Stuffing a Zebra is dan ook absoluut méér – en ánders – dan de som der delen waaruit het is opgebouwd!

​decorrespondent.nl/7401/deze-computerwetenschapper-doet-een-voorzet-voor-echt-programmeeronderwijs/588031653-1dc9ffff

(Reflections at the opening of the Platform for Ethics and Politics of Technology dd November 13th 2020)

Let me first congratulate the initiators at the University of Amsterdam with this important new Platform for the Ethics and Politics of Technology! A wonderful initiative indeed.

I am happy to be given the opportunity to pose a question to this Platform at its inauguration.

Let me start with a kind of reflection:

If there is one thing we came to realize over the past year, then it must be the realization that we live in a tremendous complex world and that we are surrounded by complex systems: from a biological cell, made of thousands of different molecules that seamlessly work together, to our society, a collection of seven billion individuals that try to live together, to the millions of computer systems that should work together.

All these complex systems display endless signatures of emerging order, disorder, self- organization and self-annihilation. Understanding, quantifying and handling this complexity is without any doubt one of the biggest scientific adventures of our time!

Now this realization that everything is connected to everything is not new, from Lao Tse (in Tao Te Ching)  to Benedictus de Spinoza (in Ethica) philosophers have pondered on this inter connectivity and asked themselves questions on how to act in that web of interwoven causes and effects.

(Benedict de Spinoza (1665): ‘Every part of Nature agrees with the whole and is associated with all other parts’.)

But then in the 20th century something special happened with the invention of the internet (and the internet of things) resulting in the ubiquitous presence of information. Suddenly the ties became more tight, the links more abundant and with an unprecedented speed we all added physical, social, economic, behavioral, emotional information into that treasure trove we now call the internet.

About the same time my field of research -complexity science- emerged in a tiny corner of the world in New Mexico where the Santa Fe Institute was pulled together by Nobel laureates, to quantitatively study these complex adaptive systems. This research is done by combining Baconian inductivism (data science, now called AI/ML) with Popperian reductionism (models and experiments) using computer simulations. This type of research into cause and effect across interwoven processes is making massive progress. Some of which is – I am happy to say- also coming from the Institute of Advanced Study here at the University of Amsterdam 12.

The consequence of all this is that, more than ever, we are able to integrate data and models and concepts from all disciplines into integrated systems that can be used to answer ‘what-if’ questions and to explore through numerical simulation the consequences of physical, infrastructural but also social and political interventions in the systems that build up our society. Think of behavioral interventions in our healthcare system, our economy, our way of handling the energy transition or the climate crisis.

The result is a kind of policy by simulation.

Which brings me to my research questions for PEPT:

Given this reflection I am pretty sure that we might soon know how to nudge people and their behavior in a way that will improve our quality of life, that might save our biodiversity, spare our scarce resources, feed the hungry, give migrants a home and build a healthy resilient society. We have come already a long way, but we now have the technology and the opportunity to move much faster and much more efficient by exploiting this interdisciplinary knowledge.

BUT, a very big but… nudging people’s behavior can be a great thing from a collective point of view but completely unethical from an individual perspective. So how to resolve this ethical disparity? And … if this nudging can happen then it will happen… that in return begs the question; Should the politics be an active agent in that process or be passive and just provide guidelines?

I sincerely hope that the PEPT initiative will consider to put these questions on their to do list, not tomorrow but now, as the need is already here and the time to act is now!

DUTCH VERSION:

De energietransitie door een complexiteitsbril
Peter M.A. Sloot, Roland Kupers en Bob van der Zwaan Institute for Advanced Study, Universiteit van Amsterdam Amsterdam, september 2019

No doubt Charles R. Darwin was right when he discovered 'the survival of the fittest', even Sharon Moalem has an excellent point when he speaks about 'the survival of the sickest' [1], but what is really shocking is what I recently discovered: 'The Survival of the Stupidest'.

Here is a way to look at it:

Our brains are constantly integrating information about the world around us. That requires neuronal activity that can not be used at the same time to to do other things, such as actually think. Multi tasking is a myth debunked time and again [2].

Deep thinking about truly complex problems require total concentration. There is no room for assessing ones environment for imminent threats. So if in the early days of the hominoids, at the dawn of homo sapiens, our species would have been engaged in deep thinking the result would be clear: the individual would simply not survive long enough to reproduce. Consequently there will be no 'deep thought' genes to be inherited [3].


 ... and then there is a Sabre-Tooth Cat lurking in the back ...









This selective process continues to date: try thinking very hard about a complex problem while crossing the road... If you start young enough doing so you can rest assured that your genes will stay unique and will not be transferred into the pool of humanity.

And because of the pre-selection that already happened over the past couple of hundred thousand years, there can only be one conclusion: We are stupid.

[1] Sharon Moalem, 'Survival of the Sickest', HarperCollins Publishers, 2007.
[2] The Myth of Multi Tasking: https://www.psychologytoday.com/blog/creativity-without-borders/201405/the-myth-multitasking
[3] Of course it is still an open debate whether 'deep thought' genes actually exist, or whether it comes about from epigenetic selection or is just adaptation or nurturing...

Astronomers of the Maya civilization and astronomers of the Babylonian civilization were brilliant in predicting astronomical events. For instance, from meticulous observations of the Sun, Moon, Venus and Jupiter they were able to predict with astonishing accuracy the 584-day cycle of Venus or the details of the celestial track of Jupiter [1]. Yet they had no clue about our heliocentric solar system, they believed that the earth was flat and they were completely ignorant of the real movement of stars and planets while being convinced that the sky was supported by four jaguars, each holding up a corner of the sky.

They were basically doing what is now called Big Data or Data Science, a very powerful way to uncover patterns in historical data. Unfortunately, data science on its own might introduce false interpretations of causality, like jaguars carrying the sky. What we need in addition to that are computational predictive models that use fundamental first principles and mechanisms that can track the system over time and allow for quantitative validation and provide pointers to novel experiments to falsify or confirm our interpretations. In other words, we need a merger of the 'inductivism' from Sir Francis Bacon (Big Data) and the 'deductivism' from Sir Carl Popper (first principle computational models) [2,3].

If we manage to turn a potential clash of these two titans of science into an integrated scientific paradigm, then the holy grail of the Scientific Method as a way 'to discover that Nature hasn’t misled you into thinking you know something you don’t actually know ' [4] will be one step closer.

See also a lecture I presented in May 2017 at the Complexity Hub in Vienna: Here.

[1] M. Ossendrijver, Science: Vol. 351, Issue 6272, pp. 482-484. (2016)
[2] P.M.A. Sloot, P. Coveney and J. Dongarra: Journal of Computational Science Vol. 1 (2010) 3–4;
[3] P.M.A. Sloot in 43 Visions for Complexity, Ed. S. Thurner, p65-66 World Scientific Publishing Co. Pte. Ltd. ISBN 978-981-3206-84-7, 2017
[4] Robert M. Pirsig, 'Zen in the art of motorcycle maintenance', 1974