One of the brainstorming tools I like to use is to take two things that are thought to be categorically different and to imagine them as a continuum. What kind of symmetry knob could be rotated gradually to turn one into the other? The ide is to purposely explore the consequences of “category errors” and question the assumption that the categories we had were the right ones to keep pure. I recently wrote about doing that to the distinction between thoughts and thinkers. Here are some ideas that come from the second part of that brainstorming strategy: seeing what implications a new perspective like that has for one’s research.
The basic idea was to recognize that typical cognitive systems (thinkers), like us, are really temporary disspative patterns within metabolic and other media. We exist for a time, while spawning off other patterns within the excitable, cognitive medium of our brain and body – i.e., thoughts, which have different degrees of solidity (fleeting thoughts, persistent thoughts, personality fragments or alters, and full on human persons). Much as we recently explored how active data might behave within a problem space (e.g., give agency to the elements of an array to be sorted, instead of a top-down boss that moves the data around), one can try to device a kind of “chemistry” for ideas or thoughts within a cognitive system that have their own dynamics. If thoughts can be (to some degree) active thinkers, one can then ask what other patterns exist which we may want to try the agential lens – to benefit from their intelligence and partial autonomy. A technical paper on this issue with Chris Fields is here.
One very interesting and important set of patterns is developmental bioelectricity: dynamic but slowly-changing spatial patterns of cellular voltage potential present in tissue. We now know many contexts in which these bioelectric patterns guide growth and form – they determine the location, size, and shape of organ-level structures in complex bodies, and are powerful targets for biomedical intervention in birth defects, cancer, and regenerative medicine. I have long suggested that, because of their evolutionary common origin and the conservation of mechanisms and algorithms between them, neural bioelectricity in the brain and developmental bioelectricity in the body can be studied in parallel ways. In other words, neuroscience isn’t really about neurons per se, but more broadly about the ways in which excitable cellular networks generate, store, and process information into adaptive behavior and a rich inner perspective.
Here are some diagrams (made by Jeremy Guay of Peregrine Creative) that I use in my talks, to get across the idea of an invariance (symmetry) between the construction of the body and the construction of the mind. They emphasize two main features that the self-construction of the mind and the self-construction of the body have in common: 1) common architectures and molecular mechanisms involving genetically-specified hardware and electrophysiological software, and 2) that by swapping control of spatial events (morphogenesis) for that of temporal events (cognition), and changing the time scale from hours to milliseconds, one can see the origin of the nervous system and conventional cognition from earlier roles of managing body anatomy. Somatic and neural bioelectric networks are basically doing the same thing: intelligently navigating a problem space (anatomical morphospace vs. familiar 3D space).
My group’s work has led to a number of discoveries from this basic realization. And much as neuroscientists record, map, and interpret brain bioelectric signals for the goal of “neural decoding”, we developed tools to read and interpret the more slowly-changing bioelectric patterns that direct development and regeneration. Here is an example of “the electric face” pattern, which presages and sets the location of the eyes, mouth, and other structures, produced by Dany S. Adams using voltage-sensitive fluorescent dye imaging of early frog craniofacial development:
Part of this neuroscience-inspired framework of understanding developmental bioelectricity as the decision-making, computational medium of morphogenetic change is the proposal that patterns like this are snapshots of the memories of the collective morphogenetic intelligence of cell groups – the proto-cognitive patterns guiding the subsequent behavior of the tissue to achieve specific outcomes in anatomical morphospace. Not in some highly metaphorical way, but literally, because they use the same molecular mechanisms (ion channels, electrical synapses) and perform many of the same functions. This has worked well for us, and led to many new experiments using the tools of neuroscience to manipulate the somatic intelligence for applications in birth defects, regeneration, and cancer.
So, what if we put two ideas – (1) patterns as agents, and (2) bioelectric patterns in living tissue – together in a new way. What if all these years, we’ve been missing an important perspective? We’ve been assuming that the body is the agent (thinker) and the bioelectric patterns were its morphological thoughts being processed (specifically, information it uses to navigate the space of possible anatomies during development, regeneration, and cancer suppression). Let’s try it a different way and turn the machine/data distinction up-side-down: what if, it’s the bioelectric patterns that are the agent, and the slowly-changing physical body is its its “memory tape”? In other words, the bioelectric pattern-driven changes we see in the cellular material (gene expression, morphogenesis, etc.) are the evidence of the agent reading and writing information into the physical world. On this view, the bioelectric states are themselves the agent (just somewhat in more subtle embodiment than the metabolic patterns embodied in arrangements of biochemical states – the conventional body), and the the body is its long-term medium or scratchpad, changing over time as the morphogenetic agent is doing the perturbative reads and writes to this scratchpad. All of the changes we observe downstream to modulating the bioelectric patterns (changes in second messenger cascades, gene expression, chromatin state, etc.) are evidence of the memory medium being accessed. As crazy as this sounds, it should be noted that a number of workers have shown evidence for the biochemical components of cells being a memory medium.
On this view, the body becomes not the ultimate target of our interventions (via the bioelectric interface) but instead, the tractable interface to what we’re really talking to, the bioelectric pattern (I guess psychiatrists would agree with this). This represents an interesting flip of perspective and likely has practical implications. We are working those out now, in the context of morphogenesis experiments in tadpole and flatworm models.
Part of that is developing better frameworks for two concepts. The first is how to think about physical laws of the universe holding patterns (a facet of the perennial “where is the pattern encoded?” question). For example, consider the significant transformation and rebuilding of the body and brain in the case of metamorphosis:
We often ask, so where is the memory stored during this process and how does it survive? Now, think about this example from the cellular automaton known as Conway’s Game of Life (GoL). In this simplified world, there is no physical glider, but it is a persistent pattern because the laws of physics (how each cell turns on/off) ensure its propagation as it moves. It has 4 life stages (morphotypes), not 2 like the caterpillar/butterfly, but nevertheless, its life cycle has the same kind of metamorphosis:
But cells in traditional GoL aren’t very smart – they only make on/off decisions based on their local neighborhood and store no representations of larger patterns. Where is the information for the persistent glider’s body which regenerates itself as it moves, every few ticks of the clock? We don’t have a great vocabulary for this yet, but it’s basically in the physics of its universe. This is a minimal model for helping us to think about how the glider pattern is using the world as its scratch-pad to maintain a consistent trajectory in morphogenetic space during its metamorphosis. This may help us think about how to design biomedical interventions for how to communicate new goals (healthy states) to the bioelectric agent, using the physical medium as the interface. Randy Beer has used this minimal model to help think about perspective-focused basal cognition.
So, is one view more right than the other? Well, when signals travel across the brain, we don’t tend to call it a memory just because they persisted long enough to get where they are going. Why not? Perhaps because the timescale of the real memory is supposed to be slower than the timescale of signals within the thinker. If the memory medium is supposed to be slower than dynamics in the agent’s processor, then the 2nd formulation above is more correct, because changes of the biochemistry and cellular structure within the body is slow while the bioelectrics are relatively faster.
Which of the two perspectives is correct – is the body the memory scratchpad of the bioelectric pattern agent, or is the bioelectric pattern the memory of the physical body agent – can we develop interventions better if we treat the body as the tape? What does that let us do? This is work in progress. But I suspect we’ll have to lean in to our polycomputing framework and realize that there is no one real answer – both perspectives have utility. In fact, it may well be that the most accurate model is that both exist at once: the body agent is using the bioelectric patterns as its memory, and simultaneously, these agential memories are doing their thing while using the physical body st their memory scratchpad. Maybe the robust problem-solving of morphogenesis is implemented by the perpetual cycle of conversation and interaction between the body-as-memory and body-as-agent. Each one models, sees, and exploits the other in whatever way it can – that’s the basis of polycomputing.
Maybe what we’re looking at in morphogenesis is a 2-way relationship: the bioelectric agent is smart and is using the body as its tape, but the body is smart too and it’s using the bioelectric patterns as its memory. It would be neat to make some simulations/models in which 2 creatures are simultaneously using each other as the lower-agency scratchpad? Is that what our brain is doing when it’s using capable neurons as a scratchpad? Are they using the “brain” at the same time – for example, to provide for them etc.? If you’re a coder interested in such things, this makes for a good ALIFE project, get in touch with me. At some point we’ll implement it ourselves in the lab and explore the implications.
But this view makes specific predictions already; for example, that aging could involve degradation of the ability to implement specific bioelectric patterns, not just the degradation of them (which we proposed here). This makes a prediction – older cells should have more trouble following bioelectric cues; that is actually observed in our recent preprint (see Figure 4). There are some very interesting experiments planned in the future, based on the inversions and blurring of the machine/data thought/thinker distinctions and testing the implications for developing new interventions in regenerative medicine.
Schematics courtesy of Jeremy Guay of Peregrine Creative.
Leave a Reply to David kolb Cancel reply