Berkeley's archetypes v Hoffman's Conscious Agents
Different approaches to the architecture of God's mind
When I first heard about Donald Hoffman’s theory of Conscious Agents (Hoffman 2019), I was quite excited, as it seemed that he was putting forward a serious account of how an idealist theory of consciousness could develop. But as more of his theory has emerged, we can see that it is heading for the same roadblocks that derail panpsychist programs.
Hoffman & Prakash (2014) offer one of the few attempts in the literature to account for the interface and architecture of consciousness within a mental monist framework. Their program is largely aligned with one described in this series of posts and my other publications (Lloyd 1999a,b; 2020a) namely to build a model of the world from a consciousness-only basis—but there are differences in the conceptualization of the mental world.
Any successful theory of idealism must tackle the basic problem of explaining how the laws of the (virtual) physical world emerge from the underlying conscious world. In the 18th Century, George Berkeley outlined an approach that was remarkably prescient of how computers render VR (Virtual Reality) environments, and how we may imagine the background consciousness of the universe renders our everyday world. I don’t want to waste time chasing labels for the background consciousness (God, Saguna Brahman, the Urgrund, the LCS (Larger Consciousness System), the metamind …), so I’ll just use the same cringeworthy word I thought up when I was a young man back in 1999, the “metamind”. Berkeley’s idea is that the metamind (which he called God) contains ‘archetypes’ that allow the metamind to render in human minds our perceptions of the manifest world. He compared this to the way a musician renders tunes from a musical score using an instrument, even though the visual symbols in the score bear no resemblance whatever to the sounds of music. That is, metamind contains compact data structures (‘archetypes’) of the observable characteristics of macroscopic bodies. It is rather like the object-oriented database of a VR system. These archetypes are then used by some mental instruments—‘rendering engines’ in computer jargon—as directives to generate our perceptions. The rendering engines are not necessarily separate: they could be implemented as ‘methods’ of the archetypes. So the physical world is understood as a simulation, coded in mental structures, and we act in it though our virtual avatars.
Any theory, such as Hoffman’s, that places consciousness in a fundamental role underpinning the physical realm faces the challenge that there is no substrate below the conscious realm that could account for features in consciousness. In Hoffman’s case, the buck stops with the Conscious Agents. This is different from, say cellular automata such as John Conway’s Game of Life. Conway’s game comprises an infinite grid of cells, each of which can full or empty, and simple rules for how the cells change from one time-step to the next. This is already a very sparse ontology, but even so the substrate of the grid provides a spatial metric against which relations and interactions are defined. When consciousness is fundamental, the ontology is even more sparse. Basic units (such as Hoffman’s Conscious Agents) are not in any prior space: they are not near to, or far from, each other. These units are neither neighbouring nor remote. There is no medium through which they can travel or communicate. Everything has to be built up from the consciousness itself. I have explored one way to approach this in my paper in the journal Entropy: Modelling Consciousness within Mental Monism: An Automata-Theoretic Approach (Lloyd, 2020a).
The bprimitive element in the model I presented is the ‘experientia’, a single instance of volition and phenomenal sensation, on top of which archetypes are assembled. In contrast, Hoffman’s primitive element is a probabilistic finite-state automaton, the conscious agent (CA), which contains a set of possible experiences and a set of possible actions. Missing from Hoffman’s account are:
An explanation of the internal mechanisms of the CA—in what concrete medium are the possible experiences and actions present?
What mechanism drives the probability distribution of the input-output relation?
How do CAs communicate with one another, given that there is no non-mental substrate that could convey signals?
What holds a CA together, to stop it leaking into other CAs?
I don’t think these are minor details, but foundational concepts. While Hoffman’s model might be valid at a high level, the lower levels also need to be fleshed out for the model to work as an account of consciousness, and that lower-level fleshing out seems to be missing.
In addition, there are certain assumptions that are different from what I have concluded in these Substack posts:
An assumption that the CAs are Markovian, that is, the state at any moment depends only on the state at the previous moment, whereas the model I have suggested is maximally non-Markovian.
An assumption that the behavior of a CA can be modelled by fixed probabilities of state transitions and outputs, while one would reasonably expect them to be time varying in complex ways, at least in the larger CAs such as human minds.
There is tension within mental monism between micro levels and macro levels. On the micro side, Hoffman has recently proposed that micro-CAs work together through abstract geometries (‘amplituhedra’) to generate spacetime and elementary quantum fields (Hoffman, 2024). On the macro side, he has long said that what we take to be physical objects (neurons, brains, planets, …) are merely icons within the MUI (Multimodal User Interface). This leads to two problems:
First, the familiar ‘combination problem’ that blocks panpsychists from forming a tenable theory. According to Hoffman, when two CAs fuse, the resulting experience set is not simply the union of the experience sets of the two merging CAs. Rather, the fused experience set contains new experiences relating to the operation of the new CA. This would be plausible if CAs were just mathematical descriptions of the functionality of some entities, but they are supposed to describe systems of actual phenomenal experience. What is it like to be a CA undergoing fusion nor fission? This question is particularly acute when it comes to our familiar human minds. In cases of local anesthesia or cortical oblation, a swathe of CAs would be knocked out, but the remainder of the sensorium is unchanged. So, it appears that the experience sets of larger entities (such as human minds) are simply the unions of the experience sets of antecedent entities. The latter is what Lloyd (2020a) has proposed: that there is a universal set of experientiae, and a private mind is formed by partitioning that universal set by closure under access.
Second, if CAs literally make up elementary physical entities (vibrations in quantum fields is the latest), then they have spatial location. This is so, even if it is the CAs that generate spacetime. This cuts against Hoffman’ long-standing claim that spacetime is just part of the ‘headset’ (his metaphor for the physical construct). If his CAs are, effectively, in spacetime, then Hoffman is going to hit the relativistic binding problem. The contents of a conscious mind have an indubitable time sequence but the neural activity of a brain that occupies a volume of space are subject to relativity of time, which gives a strong contradiction (Lloyd, 2020b).
No portals?
Finally, Hoffman does not investigate the portals that must exist between a conscious mind and its avatar in the physical construct. This is the area where we are most likely to develop falsifiable hypotheses, and develop an engineering of consciousness. Hoffman dismisses correlations between phenomenal experiences and neural activity as merely a reflection of the underlying business of CAs. While that is true, it misses the point that the brain tissue, like the rest of the physical construct (which he insists on calling a ‘headset’) is subject to rules that we have collated as laws of physics. So, while neurophysiological activity is entirely driven by what the CAs are doing, that activity is constrained to fit into the laws of physics. This immediately confronts us with the ‘portal problem’: where, in that neurophysiological activity, is there a gap in the causal chain that the CAs can exploit to exert volition in the brain? The solution that I have proposed in this paper is that consciousness acts by collapsing superpositions in DNA molecules, but Hoffman does not address the portal problem and therefore does not propose a solution.
Hoffman’s output is prolific, and I have just touched on some of the key points of comparison here. There’s a lot more to Hoffman’s theoretical account.
References
Hoffman, Donald (2008), “Conscious realism and the mind-body problem”, Mind and Matter 6(1), 87-121.
Hoffman, Donald (2010), “Sensory experiences as cryptic symbols of a multimodal user interface”, Activitas Nervosa Superior 52(3), 95-104.
Hoffman, Donald, and Chetan Prakash (2014). Objects of Consciousness. Frontiers in Psychology, 5, 1–22. https://doi.org/10.3389/fpsyg.2014.00577
Hoffman, Donald (2019), The Case Against Reality: Why Evolution Hid the Truth from Our Eyes, Allen Lane.
Hoffman, Donald, Robert Prentner, and Chetan Prakash (2023), “Fusions of consciousness”, Entropy 25(1), 129 (40 pp). https://doi.org/10.3390/e25010129
Hoffman, Donald, Chetan Prakash, and S. Chattopadhyay (2024), “Traces of consciousness”, Preprints 2024, 2024101305. https://doi.org/10.20944/preprints202410.1305.v1
Prentner, Robert, and Donald Hoffman (2024), “Interfacing Consciousness”, Frontiers in Psychology 15, 1429376. https://doi.org/10.3389/fpsyg.2024.1429376
Hoffman, Donald (2024), “Spacetime is doomed: time is an artifact”, Timing & Time Perception, 12(2), 189-191. https://doi.org/10.1163/22134468-bja10096
Lloyd, Peter B. (1999a), “Consciousness and Berkeley’s Metaphysics”, privately published.
Lloyd, Peter B. (1999b), “Paranormal Phenomena and Berkeley’s Metaphysics”, privately published.
Lloyd, P. B. (2020a). Modelling Consciousness within Mental Monism: An Automata-Theoretic Approach. Entropy, 22(6), 698–730. https://doi.org/10.3390/e22060698
Lloyd, Peter B. (2020b), TSC2020 - C3 Idealism, Panpsychism & Neutral Monism, recording: YouTube