Two stories appeared recently in the press. One reported new research suggesting that plants are conscious. The other went further and suggested that everything is conscious including stones and spoons. Of course, these studies are not quite as whacky as their media presentations suggest, but I still think that they are wrong. I am fairly confident that people are conscious and spoons are not. But why do I think this and where should we draw the line between people and spoons.
Ros and I agree that an agent is conscious if it is having subjective experiences. But why and how did consciousness evolve?
RMR: You’re asking, “what is consciousness for?” But this comprises two types of questions rolled into one. “What is it used for?” and “What evolutionary pressures produced it?” These two questions do not have the same answer.
Let’s assume that the precursor to the rich subjective experiences that human beings tell each other about was some tiny cognitive app, selected by evolution because it was useful for some minor thing. Now it turned out to open up all sorts of cognitive possibilities. What something was chosen for initially may not be the same thing as the things it finished up being useful for.
A ramekin dish has many culinary uses but its original use, possibly in seventeenth century France, is obscure. It may have evolved from other shapes by being selected for many uses rather than just one. Every time a ramekin dish is used, it is selected; it is moved from the back of the cupboard to the front. In this sense the ramekin dish is ‘for’ everything that it has been and is currently ‘used or’. Its original use may, or may not, be one of its current uses.
Evolution works in a similar way to find cognitive uses for minor mental abilities.
CDF: So, what is this tiny cognitive app from which our rich conscious experience evolved? It must evolve through some minimal form of consciousness; subjective experience without any possibility of reflecting upon it or reporting it. Wouldn’t this be consciousness without consciousness?
RMR: Yes, but that would be absurd. To explain consciousness, we need to talk about sentience: the ability to feel, to perceive, to be aware of one’s surroundings. To me, sentience is the really astonishing and mysterious aspect of consciousness. My neighbour’s cat is lying in the sun on my terrace and purring loudly. We suppose that the cat is a sentient being and that it experiences and enjoys the warmth of the sun. This form of consciousness is nicely described by Charles Kingsley in The Water-Babies (Ch3):
‘He only enjoyed it: he did not know it, or think about it.’
Next to the cat, there is a pot of parsley which is also responding to the sun. The stems are leaning towards the sun and the leaves are positioned orthogonally to the sun, maximising photosynthesis and helping the parsley to grow. But is the parsley enjoying the warmth of sun? I doubt it. Plants are not sentient, just sensitive.
CDF: But both are responding to the sun. So why do you think that the cat is sentient, but not the parsley?
RMR: The cat has eyes, nerves and a brain that have structural and physiological counterparts to those parts of the human body that, if damaged, would result in the loss or alteration of sentient experience in humans. We can even infer that the cat has colour vision that differs from ours because it has a different ratio of rods and cones in its retina. Parsley does not have this type of anatomy or any other plausible sentience-mechanism; so we presume it is not sentient, or at least, we have no reason to suppose that it might be sentient.
CDF: But can we rely on anatomy to decide whether a creature is sentient or not? Most people believe that birds and even octopuses are sentient. But they have different brains from mammals.
RMR: True, but, unlike plants, they do have brains and eyes and they can move. We can surely agree that there is a big divide between non-sentient, non-moving plants and sentient, moving animals.
CDF: Do you think that any creature that can move is sentient?
RMR: Sentience and movement may have co-evolved because both contribute to goal-directed behaviour. But invariant reflexive movements need not require sentience and can sometimes be demonstrated in humans without awareness. I think an important boundary comes between two types of modifications of movement patterns, two types of learning: classical conditioning and instrumental learning. Classical conditioning involves involuntary responses. There is no reason to assume that sentience is involved in this. The responses in instrumental learning are voluntary, and I propose that this implies sentience. The animal has to produce a spontaneous action in order to learn about its consequences.
CDF: Eva Jablonka and her colleagues reach a very similar conclusion. They suggest that the marker of sentience is the capacity for Unlimited Association Learning. This seems to be very similar to instrumental learning.
RMR: Their paper is excitingly splendid. Their evolutionary approach is very productive.Linking the evolution of sentience to reward and claiming that sentience was driven by ‘joyfulness’ is wonderful. Think of the joyful mollusc!
CDF: So instrumental learning involves voluntary actions and all sentient creatures can do it. When did this type of learning arise?
RMR: A simple example of instrumental learning is the simultaneous two-choice discrimination task and this is my choice for a test of obligatory sentience. I believe goldfish are very good at it. I guess the point when this type of learning – and sentience – first appears, would be somewhere in the primeval pond.
A simultaneous two-choice discrimination task
The monkey has to find the food under the gold knob
even though the position of the knob and the toy aeroplane varies from trial to trial.
CDF: I wonder what is it about instrumental learning or Unlimited Association Learning that entails sentience, and indeed subjective experience?
RMR: I assume this kind of learning requires making choices about perceptions. Now what are perceptions? You yourself have claimed that perceptions are an internal model of the outside world. Perception is a sentient experience and the experience is in your mind, not out there. So, sentience involves an internal model, rather than the ‘real’ world. An animal that has eyes attached to an image-making brain is making a model of the outside world. It can see; it is sentient. An animal that only has light sensitive cells is not necessarily sentient.
CDF: Here is my take on instrumental learning: it involves altering predictions. I come to have a novel expectation of what is going to happen after I press this button. Now here’s the interesting question. If you are responding to a model of the world, rather than directly to sensory stimulation, are you having a subjective experience? Jablonka talks about inputs from the body and the world being ‘mapped’ onto dynamic perception and action models and suggests that this is what is needed for phenomenal consciousness.
I wonder if sentience is the second level in a hierarchy of the control of action. At the lowest level there is a direct connection between sensory input and motor output. When sentience emerges in evolution, there is an additional level that involves action being based on a model of the world, rather than on fixed reflexes. This internal model also creates continuity across time and the potential for flexible responding.
RMR: I agree. Sentience produces continuity – the ability to join images together over the smallest moments in time. I’m sure most fully sentient animals have a ‘span of apprehension’ and ‘sense of flow of experience’ without necessarily having recollective memories or a concept of ‘then, but not now’.
CDF: We agree, a sentient creature perceives the world on the basis of a model of that world. But there is a higher level at which a creature might understand the model as being a model. There was a recent paper from Stan Dehaene and colleagues in which they called this level C2 or self-monitoring: the capacity to reflect on one’s own mental states. The lower level, which they refer to as C1, includes primary representation which we call sentience together with secondary representations, such as recollections of the past (episodic memories) and the ability to imagine alternatives, since these are also subjective experiences.
RMR: I think their paper is very clear, but it would have been nice if they had said more about non-mammalian animals – fish, birds, etc. I believe that these animals have sentience but that they can’t think about it. We don’t think about it much most of the time either. I like the idea that any goal directed behaviour involves a decision which benefits from (but is not invariably dependent on) sentience, which would explain why locomotor animals have sentience while plants do not.
CDF: I suspect that there are many animals that have sentience, but cannot entertain secondary representations.
RMR: Yes, Josef Perner suggests that human infants don’t begin to appreciate secondary representations until the second year of life. I have little difficulty believing that a chimpanzee has intention, expectation, anticipation, prediction, recollection, planning and so on. Why else would the chimpanzee be carrying so many oranges, including in his feet.
CDF: But what are the advantages of self-monitoring? I guess we now talk about a model of the model, a meta-representation.
RMR: I accept that some primates and possibly some other mammals may have some C2 monitoring of C1 events. I think this makes them self-aware i.e. self-conscious. This expands all the possibilities that had been made available by secondary representations. Even if you believe that you are seeing the real world out there, you are in fact savouring your internal experiences generated by inputs from the outside world.
CDF: There is the case of visual illusions. Here the model that we experience is not a completely reliable guide to the outside world. But if we have meta-representations and can self-monitor, then we can realise that our subjective experiences may be in error. When looking at the café wall illusion I see the horizontal lines as sloping, but I know that they actually all parallel.
We can go even further along this line of thought and start worrying about whether your perception of red is the same as my perception of red.
RMR: This is thinking about thinking, an inessential, arcane activity beloved of intellectuals.
CDF: So, we agree. Parsley is not sentient. Fish probably are. Mammals have some degree of self-awareness. But only humans can think about the subjective experience of ‘wine having an aroma of pencil shavings’.
RMR: Yes, and only humans can be so pretentious.
The Garden of Live Flowers: Sir John Tenniel 1865 Wood-engraving by Dalziel, Illustration for the second chapter of Lewis Carroll’s ‘Alice Through the Looking Glass’. From The Victorian Web
Ramekin dish: the back of cupboard
Cover of The Water-Babies: Charles Kingsley, with illustrations in colour by Warwick Goble, Macmillan 1922, from Project Gutenberg
Octopus Granulatus: Ernst Haeckel’s Kunstformen der Natur 1904, plate 54
Two-choice discrimination task: Photo courtesy of Harry Baker
Chimpanzee carrying oranges:
Café wall illusion: By Fibonacci – Own work, CC BY-SA 3.0, https://commons.wikimedia.org/w/index.php?curid=1788689