I recently published this paper, pulling together some new ideas about the nature of memory and how it applies to the nature of the Self, behavior, development, evolution, and more.
Here are some supplemental materials to go with it:
First, download here the short paper on internal voices that brought attention to a woman’s brain tumor so that it can be removed. One of my all-time favorite quotes in a paper was the last thing this voice said to her: “We are pleased to have helped you. Goodbye.” I think that understanding this kind of phenomenon (which I suspect is an unusual but not miraculous example of communication with one’s agential sub-components) is ultimately the key to a truly definitive medical roadmap. We are working on experimental approaches in this direction.
Here is a conversation about the ideas in this paper with Michael Pollan:
Here are some quotes from the paper, although you’ll have to read the whole thing for some of them to make sense:
There is a paradox which points out that if a species fails to change, it will die out, but if it changes, it likewise ceases to exist. The same issue faces all of us: if we do not change, learning and growth is impossible. If we do change, does not the current Self cease to exist, in an important sense?
When it comes to information, biology commits to optimizing salience and meaning, not fidelity, via high-level, agent-based interpretation—the gestalt, not the low-level details. Crucially, this is not just about life forms preparing for challenging external environments—it is even more about the inevitable changes of our internal parts, which are subject to mutation, aging, cancer, and hacking by other biota. More broadly, it is about the passage of time.
The essential unreliability of the biological substrate was a key driver of an architecture that is responsible for numerous fascinating capacities of the minds and bodies of life forms.
I provisionally call mnemonic improvisation the dynamic ability to re-write and remap information (e.g., memories) onto new media and new contexts, which occurs at many scales (behavioral, genetic, and physiological memories). Thus, I propose that the ability to modify and interpret non-local memories is one of the kinds of cognitive glue that reifies Selves (at all scales).
Evolution makes problem-solving agents, not solutions, and that its commitment to the reality of mutation and uncontrollable environments—to the active reinterpretation of information on the fly—is a ratchet that gave rise to intelligence and cognitive Selfhood.
I propose the hypothesis that these dynamics enabled the evolution of minimal agency in the thought forms themselves: in other words, giving up the binary dichotomy between computational machinery and passive data, and exploring the idea that certain kinds of information structures actively facilitate their transformation and remapping in ever-changing cognitively excitable media. As William James proposed, “thoughts are thinkers”.
From this perspective, memories are messages between agents separated across time— each engram is a stigmergic note left in our body by a past version of us.
I propose that the necessity for mnemonic improvisation—the active rebuilding of the content of any (proto)cognitive system—was the source of morphogenetic robustness and eventually conventional intelligence. The ability to improvise and make sense of your world in real time and the commitment to change (not just to persistence) over an allegiance to the details of a past history form a fundamental biological strategy deployed at many scales.
The process of memory and recall not only functionally generalizes (abstracts) the information (not leaves, but “food”), but most importantly, remaps it in a way that the basic relationship learned maintains its salience in a new body and environment (caterpillar -> butterfly). Stated another way, the being that radically changes does not bring with it specific memories into its new, higher-dimensional life—it brings the deep lessons of its experience in a way that it can deploy in its new embodiment.
This bowtie architecture—which forces a compression of data to a generative kernel that then has to be re-inflated and elaborated—is also a common feature of biochemical, bioelectrical, and biomechanical pathways. One of the most interesting things about compression, such as that seen in learning and generalization, is that by removing correlations, the resulting engram looks increasingly random. Because there is no “outside-text” or meta-data, the interpretation by the right side of the bowtie (or the future Self) must be creative, not only algorithmically deductive, in interpreting it in future contexts (as both the environment and body internals shift). The sense-making process of memory interpretation and the formation of models representing internal states and the external world is creative as much as it is driven by information processing and past data.
What underlying parameter is represented by the spectrum ranging from the hardwired, mosaic C. elegans to the intermediately plastic amphibia and (embryonic) mammals, and to the extreme plasticity of planaria? I propose it is the willingness to confabulate in anatomical space—a pattern extraction and completion architecture that does not take priors too seriously and assumes that it will have to develop models of itself, its problem space, and its internal and external affordances on the fly (a kind of “beginner’s mind”, emphasizing forward-looking creativity over past-constrained structure). This suggests a strategic parameter, akin to temperature in AI models, for morphogenetic/cognitive systems. In silico simulation results [176] show how this kind of process can give rise to remarkable lineages, such as planarian flatworms, which are extremely resistant to transgenesis, aging, cancer, and injury despite their incredibly noisy genome because they have fully committed to a strategy that overrides genetic details with large-scale pattern completion.
Could we blur the boundary between passive data and the active cognitive architectures that hold them—between thoughts and thinkers? For example, it has been argued that what persists are algorithms, which is a powerful way to think about active information. However, what if we go further on the continuum, beyond passive and even active data, to basal agency? Perhaps there is no principled, sharp distinction between data and algorithms, between memories and minds—but rather just a continuum of different degrees of agency between the understander and the understandant. This would also require a continuum between skills (“knowing how”) and propositional knowledge (“knowing that”). What if, in James’s words, “thoughts are thinkers” in the sense that they actively help (perhaps by cooperating and competing for opportunities or using each other as affordances in a heterarchy) cognitive systems to remap and utilize them? What if memories, which are not static details but active deep patterns, can resonate with a cognizer or even a group of cognizers (in the case of federated inference and belief-sharing) in a kind of circular causality, in which they exert some minimal agency as they shape the mind of the thinker and thus help construct the niche within which they will be utilized in subsequent time steps?
From this perspective, the continuum can range from fleeting thoughts, to persistent/intrusive thoughts, to the kinds of metastable entities experienced in tulpamancy, to dissociative and other kinds of alters, and finally to conventional full personalities (minds) that can generate all of the prior members of the hierarchy. It seems crazy to think that an agent, even a minimal kind, can be just a metastable pattern in an excitable medium—a temporarily persistent pattern. But that is what we are too—temporarily persistent, autocatalytic, dissipative patterns that self-reify our boundaries from the outside world via active inference and interpreting our environment to tell coherent stories (models) that hold us together and make us more than the sum of our parts. And on an evolutionary scale, the thoughts of a lineage mind, of which each individual creature is a hypothesis about the outside world, are definitely active agents (they are the conventional, medium-scale agents we recognize every day as behaving life forms).
Could consciousness simply be what it feels like to be in charge of constant self-construction, driven to reinterpret all available data in the service of choosing what to do next? In this sense, cognition is essentially freedom from the past; cognitive Selves could be systems that are not committed to their own past and their own memories. Paradoxically, biological Selves do not take themselves too seriously in the sense that they are not committed to a fixed set of meanings established by their prior Selflets—their freedom consists not only in actions, but in forward-looking sense-making of their own mental content. Letting go of the past Self and living life forward is a commitment to making the best of internal, not only external, information. This is in broad agreement with Solms’s idea that consciousness is palpated uncertainty about the outside world; I propose to expand this idea, with the hypothesis that consciousness is palpated uncertainty about your own memories and internal states.
Selves are simultaneously a construct in the mind of an observer(/observers), including itself, and real, causally important agents that live, suffer, die, strive, and matter. I think this also helps us in understanding what an observer is. An observer is real and significant to the extent that the content of what they observe makes a difference to them and their future behavior—they will act differently based on their interpretation of the signals they receive (unlike, for example, a telescope, or even photographic film, which forms a memory record but does not analyze in a way that links up to any cybernetic perception–action cycles). Observers interpret what they sense from their own perspective; their allegiance is to extracting meaning, not preserving accurate details. It is the mark of significant observers that they exert their agency not in seeing an event or a set of states as they are, but rather in weaving a coarse-grained compression that adaptively captures what is sensed in a process of autocrine storytelling that will be easy to exploit in future behaviors. The criterion for being an observer is that an observer is fundamentally committed to reinterpretation and meaning, not micro-scale realism. They bring their own history, perspective, biases, hedonic valence, and predictive coding strategies about what is important in sensory and interoceptive experiences. Currently, only biological beings are clearly recognized to be able to do this significantly, but these capacities do not require protoplasmic substrates or an origin through random mutations, and likely can be engineered in an endless range of novel forms of an embodied mind.
Tracking back the causality in this space of ideas suggests the following. Behavioral intelligence in 3D space evolved from morphogenetic problem-solving competencies in anatomical space. Those competencies in turn were required by the inevitability of evolutionary and physiological damage (mutations and injury), and the ratchet mechanisms of constructive neutral evolution and the paradox of robustness. The capacity for creative robustness is implemented by the polycomputing property—the ability to see the same physical process as computing and providing different functions depending on perspective. And that, in turn, comes from our multi-scale architecture: at every level, we are a collection of minimal agents that are all making sense of and hacking everything around them (their parts, their neighbors, etc.) as best as they can to fulfill their minimal but vital agendas.
Much of biology and cognitive science can thus be seen from the perspective of this fundamental paradox: “do I still exist if I change”? Creatures, whether biological or technological (or both), that resolve this Zen-like riddle do not just persist—they thrive. The inorganic world, and much of today’s engineering, is stuck in an object-centered, matter-first view. Biology embraces, and has from the start, a process ontology in which perspectives and agency are primary; thus, change is the driver of intelligence, and perspectival storytelling is a primary mechanism through which diverse minds transform and grow. I think the lesson to take from this is to embrace the dizzying freedom of breaking away from the goals and structures handed down to us from our evolutionary and personal past, and take on the responsibility of writing our own, improved somatic and mental patterns and values for the future. What engrams do you want to leave to your own future Self, and to humanity’s collective future? Despite knowing that they will not interpret them in the way you may envision now, it is still wonderous to imagine every act as a benevolent communication event to a future being.
Featured image by Midjourney.
Leave a Reply to Jaan Laaspere Cancel reply