Blackmore’s “It’s All Models”

Psychologist-philosopher Susan Blackmore claims that a fundamental mistake at the heart of consciousness studies has led to the field remaining mired in the insoluble hard problem. The mistake, according to Blackmore, was to assume that it is physical objects that are the subjects of consciousness, an assumption that leads inevitably to the explanatory gap and thus the hard problem.

How did this come about? Blackmore argues that because we have no definition of consciousness, we have relied on Nagel’s vivid question, “What is it like to be a bat?” and this question entails this fundamental assumption that the subject of ‘to be’ is the physical object—the human, rat, fish, or robot, for example. She rejects this and suggests instead that there is nothing it is like to be a bat, human, robot, etc.; there is only something it is like to be the bat’s models of a bat, the bat’s representation of itself, and the world it is flying through (Blackmore, 2025).

Conscious Models, Not Conscious Beings

Blackmore’s central claim is that all such mental models are conscious, making this theory a kind of ‘panpsychism’ of mental models. These models all have ‘something it is like’ to be them, while bats, cats, and humans do not.

Her point is this: If we ask, ‘What is it like to be a bat?’, we cannot answer. We cannot even be sure whether the bat (or dog, cat, robot, or even another human) is, or is not, conscious at all. On the other hand, if we ask, ‘What is it like to be the bat’s model of the world?’ we can answer. The answer is that if that creature is building models of its world, then there is something it is like to be whatever those models describe—whatever they say it is like.

In this view, it makes no sense to say that a human brain is conscious, or even that a whole human being is. All experience is being a mental model, whether or not there is a self-model created as an experiencer (Blackmore, 1986).

What are these models? Using a predictive processing framework, they are the active predictions that living creatures construct in order to predict and interact with their world.

These range from the simplest models that are built, for example, in early visual processing, through to intermediate and longer-lasting models of objects, places, or people, to the vast and complex self-models of humans. The subject of all our experiences is such a self-model. We are self-models in a controlled hallucination based on predictions in a Bayesian brain (Seth, 2021).

Rejecting Conscious/Unconscious Distinctions

Taking this view means rejecting a series of assumptions commonly made in consciousness research. These include the common idea that consciousness arises from or is generated by some brain processes and not others, that consciousness itself has a function that is independent of other brain functions and has evolved in addition to those, that some states or brain processes are conscious while others are not, that unconscious processes can become conscious, and that some systems or creatures are conscious while others are not.

How does Blackmore’s Predictive Mental Model theory cope with the apparent distinction between conscious and unconscious processes? Her claim is that it is our stable and complex self-models that lead us astray. There are no ‘conscious’ and ‘unconscious’ processes in a human brain, she asserts. Instead, those we call ‘conscious’ are those that are connected to, or form part of, the current self-model; those that do not, we call ‘unconscious’.

As other models come in and out of the self-model, they may appear to ‘become conscious’ or ‘enter consciousness’ (as in Global Workspace theory, for example), but this is an illusion caused by our self-centred viewpoint. This is one reason why Blackmore calls her version of illusionism ‘delusionism’, because we are systematically deluded by intuitions about our own minds (Blackmore, 2016).

The implications for research are radical. The long search for the neural correlates of consciousness can be abandoned because they do not exist. There are, obviously, neural correlates of perception, memory, thinking, and all other types of predictive model-building, but there are no correlates of something in addition to these that arise from or are generated by brain processes.

Implications for AI and Altered States

This means that research should focus not on asking where or how subjective experience arises from objective processes (as in the hard problem) but on asking what models a system is building. There is no point in asking whether a given animal, plant, or AI is conscious: questions such as ‘Is a foetus conscious?’ or ‘Is a baby conscious?’ make no sense. But we can investigate what predictive models foetuses or babies can construct.

With any such models, there is experience—the ‘what it’s like’ described by those models—but these are experiences without an experiencer. Only when there is a stable self-model are there experiences that conform to our intuitions about consciousness—that there has to be someone who is having the experience.

Blackmore’s Predictive Mental Model theory can be applied to questions about artificial intelligence. AI systems differ, and very few involve predictive processing, but this theory entails the idea that there is something it is like to be any and all models constructed artificially. However, in all current AIs, such models are fleeting, disappearing as soon as they have done their job of influencing other processes. So, although there is something it is like to be them, this is nothing like any human experience. We have bodies that interact with the world and emotions that ground the experiences and stabilize them for longer periods. So, at least for the moment, Blackmore says, AIs are not conscious in anything like the way we are.

Altered states of consciousness can be investigated by asking how predictive models, and in particular self-models, change with psychedelics, dreaming, lucid dreaming, or mystical experiences. For example, the finding that self-modelling weakens dramatically during deep sleep but strengthens with lucidity, Blackmore says, explains the sense of myself ‘waking up’ in lucid dreams.

Meditation provides an especially interesting example. Most meditation techniques involve systematically letting go of distracting thoughts while allowing simple sensations to continue. On Blackmore’s theory, what is happening is that higher-level priors are systematically dropped and temporal depth is reduced. Arguably, the process gets deeper and deeper as more and more high-level priors are let go until self dissolves, leaving experience without an experiencer, or selfless states. Long-term meditation can lead to the direct intuition that all experience is models with no separation between self and world (Blackmore & Troscianko, 2024).

Conclusion

In conclusion, Blackmore’s ‘It’s all models’ theory replaces the idea that physical objects can be conscious with the idea that only the predictive processing models they construct can be. This allows research to abandon the hunt for the NCCs, stop asking whether bats or babies are conscious, and concentrate on questions about the modelling—and especially self-modelling—that such systems create.