Sunday 4 October 2020

A Dialogical Cell Machine

I want to develop the ideas from yesterday's post about cells and networks with a closer look at cells and what a different kind of communication technology might look like that privileged cellular dynamics over individual ego. The question I left with yesterday is "What does the cell do?". I think John Torday's evolutionary biology and his First Principles of Physiology gives us a clear answer to this:

  1. A cell creates order within itself: it works against entropy, in the way Schrodinger describes
  2. A cell maintains stability with an ambiguous environment: it maintains homeostasis
  3. A cell gains energy from its environment: biologically, it gains energy through chemiosmosis, although multi-cellular organisms gain energy through digestion, photosynthesis, etc...
These principles apply at all levels of biological organisation - from cells to institutions.

Thinking about a "dialogical cell" - that is a conversation comprising a group of people, or a community, negentropy represents the ways in which people in the group organise themselves. Businesses, for example, organise themselves in functionally-differentiated units such as "accounts", "production", "marketing", etc. Academic societies organise themselves around topics and functions. This order arises through selection of particular communications - the "mutual information" of an organisation. Cognate terms serve as ways of indicating how things are organised. 

The maintenance of stability with an ambiguous environment requires that whatever happens inside a cell must represent itself to the outside environment in such a way that the cell can find a niche to survive in. Academic societies, for example, are related to one another in the topics they discuss. Each produce publications and public pronouncements of what they are doing. In so doing, they attract attention from other cells, and gain in sources of energy and support. They also contribute to the environmental conditions for their own survival - creating public communications which in time serve as an invitation to others to contribute to their survival. Such communications are rather like the "receptors" on the cell surface. 

I'm inclined to think of these external engagements as being the equivalent of an epigenetic mechanism. While the DNA of a cell might be represented by its internal organisational machinery - and a process of mutual information - the external engagement amounts to the production of epigenetic marks and the mutual redundancy between these marks and the environment producing an autocatalytic environment for the growth of the cell and the organism.  

It's a bit like a spider spinning a web - which is also a good example of redundancy. This external behaviour creates a niche for the spider, as well as transforming the environment for other organisms. That's what publications and other external communications do for dialogic cells.

In terms of energy, dialogic cells are populated by people who gain energy in other ways. However, the really important source of energy for any dialogue is new information - differences that make a difference, as Bateson would put it. Diversity in the environment of a cell is essential to the cell's survival. This is probably the biggest failing in current social networks - differences are attenuated. If there is no energy through difference, then cells are likely to eat themselves or each other. Which is pretty much what we see online. 

That's quite abstract. What about practical techy stuff? What's the functional spec for a dialogical cell machine?

What we need are:
  1. A mechanism for identifying mutual information as an organisation tool within a cell
  2. A mechanism for assisting the production of redundancy by the cell to its environment
  3. A mechanism for organising a dialogical cell such that it maximises the difference of its environment from which it can gain energy to grow
Obviously dialogue itself can identify mutual information. I saw this in the dialogues that occurred throughout the recent ANPA conference ( For example, Doug Matzke had produced a fascinating python program for doing geometric algebra, and Lou Kauffman noted the similarity between Matzke's approach and Peter Rowland's physics - "Maybe we should can put this together..." That's fine, but perhaps we can do more. In most videoed conversations now, we can produce transcripts very easily, and those transcripts can be analysed for mutual information, and references followed-up automatically, producing further mutual information. We should try this. Such analysis can be a catalyst to new forms of internal organisation in a dialogue, and it can also provide ways of managing large cells such that they might decide to split off and explore specific areas (mitosis/meiosis).

The epigenetic mechanism is more interesting because it requires some kind of system for processing and producing redundancies which might relate to external communications. What tools do we have for processing redundancy? That is precisely what machine learning does! Fed with information about internal dialogue and external dialogue, machine learning can identify external signals and generate new signals based on what is happening within the internal dialogue cell. The epigenetic mechanism is an anticipatory system. 

Left to its own devices, such a machine-learning driven mechanism will lead to confirmation bias. So a balance must be struck between the internal organisational processes and the external processes. Sometimes the epigenetic processes must generate unusual information as a way of attracting new kinds of cells from the environment. This can only happen by considering what the internal processes are doing and whether the homeostasis is "too stable". There must be a higher-level steering mechanism balancing mutual information with mutual redundancy, self-organisation and autocatalysis. 

What does all this feel like?

We start a discussion on Zoom (say). It's transcribed an analysed. New references are discovered, invitations sent out (the beginnings of epigenesis) and the discussion expands. Over time, each discussion is analysed and the result made available to its members. As internal organisational choices are made, so the analysis also identifies (through machine learning) the underlying patterns of the discussion, uses this to analyse the communications environment, and starts to support the generation of related documents (blog posts, etc), videos, etc, which are published to the outside world. Some of these attract new members... as the discussion develops, internally, the cell divides into related but separate discussions. 

I don't think this is Facebook - although there's plenty I haven't thought about yet.

No comments: