Have you ever wondered why “storytelling” is such a trendy topic? If this question bothers you and makes you uncomfortable, your perspective on human affairs and your cognitive lens is rather unusual.
Once upon a time, in the 1970s, model building gained popularity and nearly caught up with the older art of storytelling. But half a century later the popularity of model building is back to what it was in the 1960s, and according to the data analysis by Google Ngram “storytelling” has reached new heights, the word being used twice as often as the word “agile”.
The “art of storytelling” and “agile software” are strong contenders for being the catchphrase of the current millennium. Is is not surprising that the latter term was not in circulation in the 20th century, but it is perhaps somewhat surprising that the “art of storytelling” was pretty much non-existent before the 20th century.
What has happened to model building?
Unfortunately the public interface to Google Ngram only provides access to data up to 2008, but it seems that model building has given way to machine learning – and I would guess to “artificial intelligence” in recent years.
Model building is linked to the success of the scientific method. Researchers create, validate, and refine models to improve our level of understanding about various aspects of the world we live in, and to articulate their understanding in formal notations that facilitate independent critical analysis.
What is the usefulness of models that are only understandable for very few humans?
The scientific revolution undoubtedly led to a better understanding of some aspects of the world we live in by enabling humans to create more and more complex technologies. But it also created new levels of ignorance about externalities that went hand in hand with the development of new technologies, fuelled by specific economic beliefs about efficiency and abstractions such as money and markets.
In the early days of the industrial revolution modelling was concerned with understanding and mastering the physical world, resulting in progress in engineering and manufacturing. Over the last century formal model building was found to be useful in more and more disciplines, across all the natural sciences, and increasingly as well in medicine and the social sciences, especially in economics.
With 20/20 hindsight it becomes clear that there is a significant lag between model building and the identification of externalities that are created by systematically applying models to accelerate the development and roll-out of new technologies.
Humans are biased to thinking they understand more than they actually do, and this effect is further amplified by technologies such as the Internet, which connects us to an exponentially growing pool of information. New knowledge is being produced faster than ever whilst the time available to independently validate each new nugget of “knowledge” is shrinking, and whilst the human ability to learn new knowledge at best remains unchanged – if it is not compromised by information overload.
Those who engage in model building face the challenge of either diving deep into a narrow silo, to ensure and adequate level of understanding of a particular niche domain, or to restrict their activity to an attempt of modelling the dependencies between subdomains, and to coordinating the model building of domain experts across a number of silos. As a result:
- Many models are only understandable for their creators and a very small circle of collaborators.
- Each model integrator can only be effective at bridging a very limited number of silos.
- The assumptions associated with each model are only known understood locally, some of the assumptions remain tacit knowledge, and assumptions may vary significantly between the models produced by different teams.
- Many externalities escape early detection, as there is hardly anyone or any technology continuously looking for unexpected results and correlations across deep chains of dependencies between subdomains.
When the translation of new models into new applications and technologies is not adequately constrained by the level to which models can be independently validated and by application of the precautionary principle, potentially catastrophic surprises are inevitable.
Does it make sense to talk about models that are not understandable for any human?
Good models are not only useful, they are also understandable and have explanatory power – at least for a few people. Additionally, from the perspective of a mathematician, many of the most highly valued models also conform to an aesthetic sense of beauty, by surfacing a surprising degree of symmetry, by bringing non-intuitive connections into focus, and simply by their level of compactness.
Scientific model building is a balancing act between simplicity and usefulness. An overly complex model is less easy to understand and therefore less easy to integrate with complementary models, and an over-simplified model may be easy to work with, but may be so limited in its scope of applicability that it becomes useless.
What is not widely recognised beyond the mathematical community is that the so-called models generated by machine learning algorithms / artificial intelligence systems are not human understandable, for the same reasons that the physical representations of knowledge within a human brain are not understandable by humans. Making any sense of knowledge representations in a brain requires not only highly specialised scanning technologies but also non-trivial visualisation technologies – and the resulting pictures only give us a very crude and indirect understanding of what a person experiences and thinks about at any given moment.
Do correlation models without explanatory power qualify as models? There are many useful applications of machine learning, but if the learning does not result in models that are understandable, then the results of machine learning should perhaps be referred to as digital correlation maps to avoid confusion with models that are designed for human consumption. Complex correlation maps can be visualised in ways similar to the results of brain scans, and the level of insights that can be deduced from such visualisations are correspondingly limited.
It is not yet clear how to construct conscious artificial intelligence systems, i.e. systems that can not only establish correlations between data streams, but that are also capable of developing conceptual models of themselves and their environment that can be shared with and can be understood by humans. In particular current machine learning systems are not able to explain how they arrive at specific conclusions.
The limitations of machine learning highlights what is being lost by neglecting model building and by leaving modelling entirely to individual experts working in deep and narrow silos. Model validation and integration has largely been replaced with over-simplified storytelling – the goal has shifted from improving understanding to applying the tools of persuasion.
What’s the story with storytelling?
The art of storytelling is linked to the rise of marketing and persuasive writing. Edward Bernays was one of the original shapers of the logic of marketing:
Bernays’ vision was of a utopian society in which individuals’ dangerous libidinal energies, the psychic and emotional energy associated with instinctual biological drives that Bernays viewed as inherently dangerous given his observation of societies like the Germans under Hitler, could be harnessed and channelled by a corporate elite for economic benefit. Through the use of mass production, big business could fulfil the cravings of what Bernays saw as the inherently irrational and desire-driven masses, simultaneously securing the niche of a mass production economy (even in peacetime), as well as sating what he considered to be dangerous animal urges that threatened to tear society apart if left unquelled.
Bernays touted the idea that the “masses” are driven by factors outside their conscious understanding, and therefore that their minds can and should be manipulated by the capable few. “Intelligent men must realize that propaganda is the modern instrument by which they can fight for productive ends and help to bring order out of chaos.”
The conscious and intelligent manipulation of the organized habits and opinions of the masses is an important element in democratic society. Those who manipulate this unseen mechanism of society constitute an invisible government which is the true ruling power of our country. …In almost every act of our daily lives, whether in the sphere of politics or business, in our social conduct or our ethical thinking, we are dominated by the relatively small number of persons…who understand the mental processes and social patterns of the masses. It is they who pull the wires which control the public mind.
Propaganda was portrayed as the only alternative to chaos.
The purpose of storytelling is the propagation of beliefs and emotions.
What is the usefulness of stories if they do nothing to improve our level of understanding of the world we live in?
Sure, if stories help to increase the number of shared beliefs within a group, then the people involved may understand more about the motivations and behaviours of the others within the group. But at the same time, in the absence of building improved models about the non-social world, the behaviour of the group easily drifts into more and more abstract realms of social games, making the group increasingly blind to the effects of their behaviours on outsiders and on the non-social world.
Stories are appealing and hold persuasive potential because of their role in cultural transmission is the result of gene-culture co-evolution in tandem with the human capability for symbolic thought and spoken language. In human culture stories are involved in two functions:
- Transmission of beliefs that are useful for the members of a group. Shared beliefs are the catalyst for improved collaboration.
- Deception in order to protect or gain social status within a group or between groups. In the framework of contemporary competitive economic ideology deception is often referred to as marketing.
Storytelling thus is a key element of cultural evolution. Unfortunately cultural evolution fuelled by storytelling is a terribly slow form of learning for societies, even though storytelling is an impressively fast way for transmitting beliefs to other individuals. Not entirely surprisingly some studies find the prevalence of psychopathic traits in the upper echelons of the corporate world to be between 3% and 21%, much higher than the 1% prevalence in the general population.
Storytelling with the intent of deception enables individuals to reap short-term benefits for themselves to the longer-term detriment of society
The extent to which deceptive storytelling is tolerated is influenced by cultural norms, by the effectiveness of institutions and technologies entrusted with the enforcement of cultural norms, and the level of social inequality within a society. The work of the disciples of Edward Berneys ensured that deceptive storytelling has become a highly respected and valued skill.
However, simply focusing on minimising deception is no fix for all the weaknesses of storytelling. When a society with highly effective norm enforcement insists on rules and behavioural patterns that create environmental or social externalities, some of which may be invisible from within the cultural framework, deception can become a vital tool for those who suffer as a result of the externalities.
Furthermore, even in the absence of intentional deception, the maintenance, transmission, and uncritical adoption of beliefs via storytelling can easily become problematic if beliefs held in relation to the physical and living world are simply wrong. For example some people and cultures continue to hold scientifically untenable beliefs about the causes of specific diseases.
All political and economic ideologies rely on storytelling
Human societies are complex adaptive systems that can’t be described by any simple model. More precisely, it is not possible to develop long-range and detailed predictive models for social and economic behaviour. However, in a similar way that extensive sensor networks and modern computing technology allows the development of useful short-range weather forecasts, it is possible to use social and economic data to look for externalities and attempts of corruption.
Nothing stands in the way of monitoring the results of significant social and economic changes with a level of diligence that is comparable to the diligence expected from researchers when conducting scientific experiments in the medical field. Of course the pharmaceutical industry also has a reputation for colourful storytelling, and the healthcare sector is not spared from ethical corruption and the tools of marketing. But at least the healthcare sector is heavily regulated, academic research is an integral part of the sector, and independent validation of results is part of the certification process for all new products and treatments.
One has to wonder why economic and social policies are not subject to a comparable level of independent oversight. The model of governance in modern democracies typically includes a separation of power between legislature, executive, and judiciary, but the question is whether effective separation of power can be maintained over decades and centuries.
Human societies and social structures are far from static. Concepts such as the nation state are only a couple of hundred years old and the lifespan of economic bubbles and the structures created by within such bubbles is measured in years rather than centuries. And yet, many people and institutions are incapable of considering possible economic or social arrangements that lie outside consumerism and the cultural norms that currently dominate within a particular nation state. Cultural inertia is beneficial for societies whenever the environment in which they are embedded is highly stable, but it becomes problematic when the environment is undergoing rapid change.
Historically a rapidly changing environment used to be associated with local wars or local natural disasters such as extended periods of draughts or earthquakes. The industrial revolution has significantly shifted the main triggers of rapid change:
- Improvements in technology, hygiene and medicine have facilitated significant population growth and ushered in a new geological era – the anthropocene, human activity is changing the physical environment faster than ever before
- Machine powered technology has enabled wars of unprecedented scale, speed, and levels of destructiveness
- The paradigm of growth based economics fuelled by interest bearing debt and aggressive marketing dominates on all continents in most societies, and facilitates global economic bubbles
- Carbon emissions and other physical externalities of modern economic activity have no physical boundaries
Given this context it is extremely tempting for professional politicians within government and corporations to subscribe to the elitist logic of Edward Bernays and to exploit storytelling for local or personal gains. An alien observer of human societies would probably be amazed that some humans (and large organisations) are given a platform for virtually unlimited storytelling at a scale that affects billions and hundreds of millions people, and that delusional and misleading stories are let lose on the population of a species that is the local champion of cultural transmission on this planet.
Within growth based economics the effectiveness of marketing can never be good enough. Desperate corporations are hoping machine learning algorithms can take storytelling to yet another level. High frequency trading is one example of “successful” automated marketing, where algorithms try to trick each other into believing stories that are beyond human comprehension.
End of story?
If we continue to believe that the world is shaped exclusively by human delusions, then the human story may come to a fairly unspectacular end rather soon. It also won’t help us if we focus on building technologies that provide even more powerful delusions.
If there is anything that has led to significant improvements in human well-being and life expectancy in the last thousand years it would undoubtedly have to be model building and the scientific method. The power tools of systematic experimentation and modelling facilitated much of what we call progress but they also facilitated dangerous social games at a planetary scale.
Just as medical science no longer relies on unsubstantiated stories, the stories that we tell each other in business, government, and academic administration need to be subjected to critical analysis, and the public needs to be made aware of the evidence (or lack thereof) that underpins the claims of politicians and executives in the corporate world, so that experiments are clearly identified, and most importantly, that experiments are carefully monitored and subjected to independent review before being sold as solutions.
In this context lessons can be learned from the fast moving world of digital technology. On the positive side the software development community is acutely aware of the need to conduct experiments, on the negative side, outside a few life critical industries, the lack of rigour when conducting experiments in the development and deployment of new software solutions is embarrassing. In the software development community conducting multiple independent experiments is generally considered a waste of time, and the interests of financial investors determine the kinds of “solutions” that receive funding:
All human artefacts are technology. But beware of anybody who uses this term. Like “maturity” and “reality” and “progress”, the word “technology” has an agenda for your behaviour: usually what is being referred to as “technology” is something that somebody wants you to submit to.
“Technology” often implicitly refers to something you are expected to turn over to “the guys who understand it.” This is actually almost always a political move. Somebody wants you to give certain things to them to design and decide. Perhaps you should, but perhaps not.
– Ted Nelson, a pioneer of information technology, philosopher, and sociologist who coined the terms hypertext and hypermedia in 1963.
The software industry is an interesting economic subsystem for observing human social behaviour at large scale. Today this sector is interwoven with virtually all other economic subsystems and even with the most common tools that we use for communicating with each other.
David Graeber has analysed the phenomenon of “bullshit jobs” in detail.
“In the year 1930, John Maynard Keynes predicted that technology would have advanced sufficiently by century’s end that countries like Great Britain or the United States would achieve a 15-hour work week. There’s every reason to believe he was right. In technological terms, we are quite capable of this. And yet it didn’t happen. Instead, technology has been marshalled, if anything, to figure out ways to make us all work more. In order to achieve this, jobs have had to be created that are, effectively, pointless. Huge swathes of people, in Europe and North America in particular, spend their entire working lives performing tasks they secretly believe do not really need to be performed. The moral and spiritual damage that comes from this situation is profound. It is a scar across our collective soul. Yet virtually no one talks about it. …”
Silicon Valley innovation pop-culture?
Students of software engineering and computer science are often attracted by the idea of “innovation” and by the prospect of exciting creative work, contributing to the development of new services and products. The typical reality of software development has very little if anything to do with innovation and much more with building tools that support David Graeber’s “bullshit jobs” and Edward Bernays’ elitist “utopia” of conscious manipulation of the habits and opinions of the masses by a small number of “leaders” suffering from narcissistic personality disorder.
The culture within the software development community is shaped much less by mathematics and scientific knowledge about the physical world than by the psychology of persuasion – and an anaemic conception of innovation based on social popularity and design principles that encourage planned obsolescence. A few years ago Alan Kay, a pioneer of object-oriented programming and windowing graphical user interface design observed:
It used to be the case that people were admonished to “not re-invent the wheel”. We now live in an age that spends a lot of time “reinventing the flat tire!”
The flat tires come from the reinventors often not being in the same league as the original inventors. This is a symptom of a “pop culture” where identity and participation are much more important than progress. … In the US we are now embedded in a pop culture that has progressed far enough to seriously hurt places that hold “developed cultures”. This pervasiveness makes it hard to see anything else, and certainly makes it difficult for those who care what others think to put much value on anything but pop culture norms.
Mainstream software development practices are geared towards dealing with the characteristics of big ball of mud architectures and the reality of the curse of software maintenance.
Do we need a better language for model building?
Making model building accessible to a wider audience may require developing a cognitively simple visual language for articulating resource and information flows in living and economic systems in a format that is not influenced by any particular economic ideology.
Many of the languages of mathematics already make use of visual concept graphs. Digital devices open up the possibility of highly visual languages and user interfaces that enable everyone to create concept graphs that are formal in a mathematical sense, understandable for humans, and easily processable by software tools. The only formal foundations needed implementing such a visual language system are axioms from model theory, category theory, and domain theory.
In terms of usability, a formal software-mediated visual language system that takes into consideration human cognitive limits has the potential to:
- Improve the speed and quality of knowledge transfer between human domain experts
- Improve the speed and quality of knowledge transfer between human domain experts and software tools
- Facilitate innovative approaches to extracting human understandable semantics from informal textual artefacts, in a format that is easily processable by software tools
- Facilitate innovative approaches to unsupervised machine learning that deliver results in a format that is compatible with familiar representations used by human domain experts, enabling the construction of knowledge repositories capable of receiving inputs from:
- human domain experts
- informal textual sources of human knowledge
- machine learning systems
All scientists, engineers, and technologists are familiar with a language that is more expressive and less ambiguous than spoken and written language. The language of concept graphs with highly domain and context-specific iconography regularly appears on white boards whenever two or more people from different disciplines engage in collaborative problem solving. Such languages can easily be formalised mathematically and can be used in conjunction with rigorous validation by example / experiments.
Model building and digital correlation maps can go hand in hand
Machine learning need not result in opaque systems that are as difficult to understand as humans, and a formal visual language may represent the biggest breakthrough for improving the understanding between humans since the development of spoken language.
… And storytelling and social transmission need not result in a never ending sequence of psychopathic social games if we get into the habit of explicitly tagging all stories with the available supporting evidence, so that untested ideas and attempts of corruption become easier to identify.
In all domains where decisions and actions may have significant impact on others and on the environment we live in, adopting a more autistic mindset in relation to human stories may improve human decision making. In the Asch conformity experiment, autists were found to resist changing their spontaneous judgement to an array of graphic lines despite social pressure to change by conforming to the erroneous judgement of an authoritative confederate.
Mathematics – the language of explanation and validation
Paul Lockhart describes mathematics as the art of explanation. He is correct. Mathematical proofs are the one type of storytelling that is committed to being entirely open regarding all assumptions and to the systematically exploring all the possible implications of specific sets of assumptions. Foundational mathematical assumptions are usually refereed to as axioms.
Formal proofs are parametrised formal stories (sequences of reasoning steps) that explore the possibilities of entire families of stories and their implications. Mathematical beauty is achieved when a complex family of stories can be described by a small elegant formal statement. Complexity does not melt away accidentally. It is distilled down to the its essence by finding a “natural language” (or “model”) for the problem space represented by a family of formal stories.
A useful model encapsulates all relevant commonalities of the problem space – it provides an explanation that is understandable for anyone who is able to follow the reasoning steps leading to the model.
The more parameters and relationships between parameters come into play, the more difficult it typically is to uncover cognitively simple models that shed new light onto a particular problem space and the underlying assumptions. If a particular set of formal assumptions is found to have a correspondence in the physical or living world, the potential for positive and negative technological innovation can be profound.
Whether the positive or negative potential prevails is determined by the motivations, political moves, and stories told by those who claim credit for innovation.
Any hope of progress beyond stories?
From within a large organisation culture is often perceived as being static or very slow moving locally, and changes in the environment are being perceived as being dynamic and fast moving. This is an illusion. It is easy to lose sight of the bigger picture.
Outside the context of “work” the people within a large organisation are part of many other groups and part of the rapidly evolving “external” context. The larger an organisation, the greater the inertia of the internal organisational culture, and the faster this culture disconnects from the external cultural identities of employees, customers, and suppliers.
The resulting cognitive dissonance manifests itself in terms of low levels of employee engagement, high levels of mental illness, and the increasingly short life expectancy of large corporations. Group identities and concepts such as intelligence and success are cultural constructs that are subject to evolutionary pressure and phase transitions.
Marketing may well become a taboo in the not-too-distant future
Over the next few weeks, to my knowledge, there are at least four dedicated conferences on the topics of redefining intelligence, new economics, and cultural evolution.
Conference on Interdisciplinary Innovation and Collaboration
Melbourne, Australia – 2 September 2017
Auckland, New Zealand – 16 September 2017
These events are part of the quarterly CIIC unconference series, addressing challenges that go beyond the established framework of research in industry, government and academia. The workshops in September will build on the results from earlier workshops to explore the essence of humanity and how to construct organisations that perform a valuable function in the living world.
The historic record of societies and large organisations being aware of the limitations of their culture is highly unimpressive. Redefining intelligence is our chance to break out of self-destructive patterns of behaviour. It is a first step towards a better understanding of the positive and negative human potential within the ecological context of the planet.
More information on CIIC and the theme for the upcoming unconference:
Thanks to Pete Rive and Arthur Shelley at AUT and RMIT for providing CIIC with superb venues!
New Economy Conference
Brisbane, Australia – 1-3 September 2017
Building on the inaugural 2016 conference held in Sydney, the 2017 gathering invites people to come together to share stories of success, address challenges and join the broader movement so we can continue working together to build a ‘new’ economic system. The 2017 New Economy Conference will bring together hundreds of people and organisations to launch powerful new collective strategies for creating positive social and economic change, to achieve long term, liveable economies that fit within the productive capacity of a healthy environment.
More information on NENA and the Building the New Economy conference:
Thanks to Donnie Maclurcan and Tirrania Suhood for making me aware of this event!
Inaugural Cultural Evolution Society Conference
Jena, Germany – 13 – 15 September 2017
The Cultural Evolution Society supports evolutionary approaches to culture in humans and other animals. The society welcomes all who share this fundamental interest, including in the pursuit of basic research, teaching, or applied work. We are committed to fostering an integrative interdisciplinary community spanning traditional academic boundaries from across the social, psychological, and biological sciences, and including archaeology, computer science, economics, history, linguistics, mathematics, philosophy and religious studies. We also welcome practitioners from applied fields such as medicine and public health, psychiatry, community development, international relations, the agricultural sciences, and the sciences of past and present environmental change.
More information on CES and the related conference:
Thanks to Joe Brewer and his team for coordinating this unique event!