… and the language of human speech is overrated.

Given that human children learn to use spoken language to attach labels to mental representations very early on, and given that much of human communication is based on spoken and written language, it is tempting to perceive human language as our main thinking and reasoning tool.

The more we learn about the reasoning abilities of non-human animals, the more doubt is cast on the position of human language as the ultimate thinking tool.

Many philosophers have used language to write about their mental models of the world, and yet, when we read their explanations, to what extent can we be confident that the mental models that we reconstruct from a written text resemble something similar to what the writer had in mind?

Argumentative writers and speakers of human languages easily forget is that every mind is different, has made unique experiences, and therefore creates unique mental models of the world. Arguing about which mental models are “better” or “correct” has entertainment value for some, but it distracts from the fact that mental models can’t be debated – the debates are always about partial mental models that have been serialised into human language and that are interpreted out of context, without the substrate of all the unique experiences that gave rise to the shared models.

The atoms of the “language” of thought

Human mental models have been around for much longer than human language. To understand the core mechanisms of human reasoning and thinking, and to appreciate the dangerous limitations of human language, we need to step back in time and look at how language evolved from a biological perspective.

Here is a synopsis of the thinking tools that predate human language:

1. Humans and a some other animals are capable of shared attention. I can look at something and detect that another animal is looking at the same thing, and I understand that we are both seeing the same thing, whilst realising that we may have wildly different perspectives on the thing (associations with past experiences) that we see. Someone who has never seen or heard of a gun may not know that it can kill. I can also observe two people who are looking at some object, and I understand that their minds are focused on that object.
2. Beyond awareness of shared attention humans have evolved limbs that allow us to point to things, to further disambiguate and make it more obvious what we are focusing on.
3. Humans and other animals create mental representations (= models) of the things we interact with.
4. Furthermore humans and some animals can identify commonalities between things (abstract/generalise) and create mental models of groups of similar things (= categories).
   • … and can identify spatial relationships between things (containment and connectors) and create mental models of these relationships (= graphs).
   • … and can identify changes over time (movement of things) and create mental models of patterns of movements (= operations).
5. Humans and perhaps also some animals can apply their pattern recognition and abstraction abilities to operations, leading to mental representations that contain abstract operations.
6. Humans and perhaps also some animals rely on their mental models to conduct extensive simulations to predict events and arrive at decisions. In some domains this happens subconsciously and very fast, and in other domains we are capable of slower and deliberate conscious simulations.

We and other animals can do all of these things without talking! No spoken or written language is required. Mental models and reasoning clearly came first. Human language came second.

Achieving shared understanding

Prior to spoken or written language communication via shared attention and pointing was the was way of establishing shared understanding, and such shared understanding related to very down-to-earth representations.

We probably developed the technique of validation by instantiation (pointing to multiple examples) long before language, to establishing a reasonable level of shared understanding of life-critical abstract concepts. Perhaps it became convenient to associate some of these abstract concepts with reasonably distinct kinds of grunts, and this then eventually led to how we speak.

Some philosophers – and software developers – are so infatuated with language and syntax that they can become blind to how the physical world is shaping our mental worlds.

Pre-language humans communicated within a highly local context in space and time. The things being referenced were “close at hand”. It was reasonable for people to assume that others understood what they referred to. The risks for misunderstandings were limited.

Spoken language

As soon as spoken language entered our world, initially as a serialisation format for communicating simple references to things within our local context, things started to get messy. We started to reference abstract things, references to references, and experiences that occurred many years ago. From that time onwards I suspect the number of misunderstandings in communication grew exponentially.

Theory of mind has likely evolved in tandem with language, it allowed us to create rough and speculative models of what might go on in another mind. But since people could not visit the past of other people, this lead humans down the path of extensive social delusion, where they started to assume that they understood each other much better than they actually did.

Validation by instantiation would now only come into play when harsh reality pointed people to concrete misunderstandings. Something like the 80 / 20 rule will have been good enough for language to be a useful and viable tool. Making correct assumptions 80% of the time was good enough for day to day life.

Written language

The seeds for storytelling had been sowed. The first human hive minds emerged.

Written language made things even worse in terms of the scope of the social delusion. People had opportunities to “read” large volumes of information out of context in space and time. Many of the written words of old and distant texts seemed familiar – as needed with the help of a translator (another source of potential misunderstandings), and people started importing many thousands of references to very unfamiliar abstractions into their mental models on top of their first hand experiences.

We all know that human imagination knows no/few limits, but at the same time we like to believe that we “understand” what others have written, without necessarily realising the contradiction. The human tendency to believe in the validity of our imagination after hearing or reading a story allowed storytelling and belief systems to rise to new heights.

Human culture increasingly became defined by myths residing within hive minds. At that stage a few people started scratching their heads about weird human behaviours and the beliefs that underpinned the observed behaviours. In today’s society, within the pathology paradigm of Western medicine, such people would be labelled autistic.

Mathematics

At all times throughout human history a few people would have realised that human language has severe limitations in terms of ambiguity and precision. This led to the development of number systems and mathematical tools. The industrial revolution and the computer revolution would not have been possible without reliable mathematical tools.

Given the limitations of human languages, it is perhaps not entirely surprising that modern foundations of mathematics take us back to core concepts that pre-date human language – to the atoms of the language of thought.

Model theory expresses the biological foundations of human mental models in a formal symbol system. And since mathematicians were at work, they used recursion to made sure that this symbol system is fit for representing and reasoning about all other kinds of symbol systems and syntaxes.

Denotational semantics is based on the simple observation that with our current technologies we can easily construct sufficiently unique symbolic tokens that allow us to abstract the use of human understandable symbol systems into corresponding machine readable symbol systems and to store the references between the two.

Category theory is a thinking tool for articulating large scale patterns and establishing semantic equivalences between different domains, it does not involve any concrete symbol systems. We perform such semantic calculations in our minds all the time, mostly subconsciously.

Visualising the language of thought

… and integrating it with digital computers:

In the Cell technology implementation of mathematical foundations we use unique tokens (semantic identities) to denote symbols and symbolic language systems.

We recognise that all language (and especially meaning/semantics) is created and shaped by individual symbol system users. All conceptual references are associated with exactly one agent, accounting for the fact that we all have unique mental models and that the cultural hive mind is a social delusion.

We rely on validation via instantiation to establish a reasonable level of shared understanding between agents (human and software), and on information quality (IQ) logic in combination with Bayesian logic to estimate the level of shared understanding in relation to specific contexts/domains.

Design of trustworthy protocols

… for interacting with humans and machines:

The foundational and irreplaceable role of the language of thought for reasoning and decision making explains why model building is so important. With spoken and written language and digital “storytelling on steroids” enabled by zero marginal cost communication modern humans have painted themselves into a dangerous corner – the goal of communication has shifted from knowledge validation and achieving shared understanding towards the transmission of seductively simplistic beliefs.

Mathematics, the arts, and music are human scale tools for communicating the essence of complex patterns of mental states (knowledge, feelings, and awareness of agency and motivations) that don’t survive simplistic attempts of serialisation and de-serialisation via stories. The outputs of mathematics, the arts, and music are highly generative, they can’t be described in any simple story. Instead they open up and invite a multitude of complementary interpretations.

The interpretation or application of mathematical theories has become a critical part of a growing number of knowledge intensive disciplines. The essence of the scientific method is the combination of the atoms of the language of thought with the technique of validation via instantiation. The arts and music are essential complementary communication and exploration tools for feelings, agency, and motivations.

Of course creative storytelling also allows for a multitude of interpretations, but whenever storytelling and related tools of persuasion are used to transmit and replicate beliefs, as is usually the case in politics and marketing, interpretation out of context becomes problematic, and critical validation becomes essential to minimise misunderstandings and attempts at deception.

If we value the creation of cultures of thinking, then the risks of deceptive storytelling need to be acknowledged, and exploration and critical validation of knowledge, feelings, agency, and motivations must be encouraged.

Join us at the next CIIC unconference on 2 December to explore the design of new types of dialogues and interaction patterns specifically for knowledge validation and trust building in our increasingly interdisciplinary world.

---

Venues
Dates and times