Let’s look at each of these categories in turn.
- Neural correlates of consciousness and “signatures of
consciousness”
The neural correlates of consciousness (NCC) have received the most
scientific attention as a means for assessing the presence and type of
consciousness. This term seems to have been coined by Crick and Koch as
part of their quest to determine what parts of the brain are necessary
and sufficient for conscious experience (Crick and Koch 1990; Koch 2004;
Koch 2019).
When, for example, determining whether a vegetative patient is conscious
in any way, we can and do examine the neural correlates of consciousness
only, since there aren’t any behaviors to observe and no creative
products either. Various researchers have
proposed tests for
cognition and consciousness in coma and vegetative patients.
What’s physically going on in the brain in such cases? Neuroimaging
tools such as EEG, MEG, fMRI and TMS (transcranial magnetic
stimulation), each with their own strengths and weaknesses, are able to
provide information on activity happening within the brain even in coma
and vegetative
patients.
Dehaene and his colleagues have identified four
“signatures
of consciousness,” which extend the idea of neural correlates of
consciousness to more specific aspects of brain activity that are
necessary for conscious awareness, rather than only being correlated
with consciousness (this distinction, however, may be a misunderstanding
of Koch’s definition of NCC since Koch’s definition is just this).
Dehaene focuses on what’s known as the “P3 wave,” which is the
positive electric field (measured with EEG) that occurs about 300
milliseconds in the dorsolateral cortex, after a sufficiently strong
stimulus, as the single most important signature of consciousness in
humans. In cortex where stimuli don’t result in a P3 wave, Dehaene et
al. suggest that this activity does not reach the level of
consciousness. In tests of vegetative and minimally conscious patients,
he and his colleagues have
successfully predicted
which patients are most likely to regain more normal states of
consciousness.
Sid Kouider, another French neuroscientist, has
examined very young
babies in order to assess the likelihood of them being conscious, using
a similar set of signatures of consciousness as Dehaene and his
colleagues. He concludes (it would be surprising if he hadn’t come to
this conclusion) that even newborns are conscious in various complex
ways.
Casarotto et al. 2015 measure a Perturbational Complexity Index (PCI) as
a simpler proxy for consciousness conceived as integrated information.
Tononi is a coauthor on this paper and Tononi’s Integrated Information
Theory (IIT) is widely considered to be one of the more convincing
extant theories of consciousness. IIT suggests that integrated
information is consciousness, though in its most recent formulation
consciousness is instead considered to be identical to the
maximally-integrated causal ___ (“MICE”), see Koch 2019;
Oizumi, et al. 2014.
Tononi and his colleagues acknowledge that measuring MICE or integrated
information (\(\Phi\)) in any biologically-complex context is extremely
difficult. The PCI was developed as a more tractable means for measuring
mammalian brain activity and relies on quantification of electrical
activity with sophisticated EEG during transcranial magnetic stimulation
(TMS) of the brain. PCI is a measure of elasticity of neurons and neuron
complexes under perturbation by TMS. It is thought that the less elastic
a neuron is the more it is firing and thus contributing to consciousness
(Casarotto et al. 2015): “PCI offers a reliable, independently
validated stratification of unresponsive patients that has important
physiopathological and therapeutic implications. [We found that] the
high-PCI subgroup of VS patients may retain a capacity for consciousness
that is not expressed in behavior.”
The PCI measures interconnectedness between different parts of the
brain, and activity within those information pathways: “PCI directly
gauges the ability of many functionally specialized modules of the
thalamocortical system (differentiation) to interact rapidly and
effectively (integration), thus producing complex patterns of
activity.” (Id.)
Some researchers, such as Koch 2019, view the PCI measure as an
effective but far from perfect psychometer even in its current state.
Koch elaborates on his views on PCI in his 2019 book, The Feeling
Of Life Itself , worth quoting at length (Koch 2019, p. 104):
Zap-and-zip [the colloquial name for the PCI technique] does not
measure integrated information. While the PCI index is motivated by IIT,
it crudely estimates differentiation and integration. Pace IIT, a
genuine consciousness meter should measure\(\Phi\)max. Such a \(\Phi\)-meter or phi-meter must
causally probe the NCC at the relevant level of spatiotemporal
granularity that maximizes integrated information. This can be assessed
empirically.
A true phi-meter should reflect the waxing and waning of experience
during wakefulness and sleep, how consciousness increases in children
and teenagers until it reaches its zenith in mature adults with a highly
developed sense of self, with, perhaps, an absolute maximum on long-term
meditators, before it begins its inevitable decline with age. Such a
device would generalize across species, whether or not they have a
cortex, or, indeed, any sort of sophisticated nervous system. For now,
we are far from such a tool. In the interim, let us celebrate this
milestone in the millennia-old mind-body problem.
In sum, PCI and “zap-and-zip” arguably represent a high-water mark for
current efforts to measure complex consciousness, and yet PCI remains a
relatively crude measure.
Looking to the evolution of consciousness, and what it takes to enjoy
consciousness in a form resembling human consciousness, Feinberg and
Mallatt 2016 presents a theory of consciousness, which they call the
theory of “neurobiological naturalism. They argue the following
features are necessary for the presence of consciousness: 1) living
processes; 2) multicellularity with a nervous system and a basic core
brain; 3) “numerous and neurobiologically unique special
neurobiological features of complex nervous systems.” The authors
conclude that consciousness is a relatively recent phenomenon, emerging
from the explosion of animal complexity in the Cambrian Explosion around
540 million years ago. This theory is so obviously contingent and
tautological, offering no explanation as to why these specific features
are necessary for consciousness, that I shall not dwell on it further.
- Behavioral correlates of consciousness
The behavioral correlates of consciousness (BCC) are a more intuitive
and more common manner of assessing the presence and type of
consciousness in everyday life and even in a clinical setting. When
measuring NCC is not an option, we need to consider behavioral
correlates as clues for the presence and type of consciousness. As with
measures of the NCC there is no established methodology for using BCC as
a measure(s) of consciousness.
At the most general level, we as humans assess the presence and type of
consciousness present in other humans multiple times per day through
conversations with other humans. Speech and other forms of direct
communication are a type of BCC. Because each us uses speech to convey
our own thoughts and feelings too often to keep track of we also accept
that other humans expressing complex, or even simple, thoughts and
feelings through the use of speech are in fact conscious in ways very
similar to ourselves. This statement is not revelatory but it is
important to state as a baseline approach for assessing the presence of
consciousness.
In the AI context and simulation of consciousness – independent of
deeper questions about whether AIs can ever be truly conscious rather
than merely simulating it, the well-known Turing Test (described as
“the imitation game” by Turing himself, Turing 1950) is meant to allow
humans to judge the degree of thinking ability (what we would today
describe as consciousness), if any, in an AI. The Turing Test was
originally proposed to allow human participants to make a specific
judgment based on responses from the test subject about the nature of
the subject: could it really think? Was it a human or a machine? The
Turing Test is a way of using BCC to assess the presence and type of
consciousness in AIs through direct communications.
The BCC encompass far more than forms of communication, however.
Cats cannot communicate their states of consciousness with words. The
brain architecture in cats is quite different from humans and they have
very minimal prefrontal cortex, which is thought to be the center of
many higher-order
activities of the human brain. But is prefrontal cortex necessary for
consciousness?
Cat behavior is complex and readily mappable onto human behavior in many
ways. The fact that cats purr, flex their toes and snuggle when petted,
in similar ways to humans demonstrating pleasure when physically
stimulated (minus the purrs, of course), meow loudly for food when
hungry, and stop meowing when fed, demonstrate curiosity or fear about
other cats or humans with various types of body language, and many other
behaviors that we can easily observe ourselves if we have cats as pets,
is fairly convincing evidence, for most of us, that cats are indeed
conscious and have a rich emotional life.
LeDoux 2019 argues for a conservative approach to BCC. He argues that
the default position for making any determinations about the presence of
consciousness should, if possible, explain behaviors without inferring a
role for consciousness. I advocate a less conservative view below, which
is part of the more recent progressive trend toward treating the study
of animal minds in the same or similar ways as we do human minds. This
trend is itself based primarily on the obvious strong kinship and
similarities between humans and other animals, as well as the
well-established fact that biological change occurs slowly and
incrementally, abhorring “jumps.”
Donald R. Griffin has examined the nature of animal minds in various
works, including his 2001 book, Animal Minds: Beyond Cognition to
Consciousness , in which he states (Griffin 2001, p. xiii): “In a sharp
break with the traditional conviction that the mental experiences of
animals cannot be studied scientifically, some of us have begun to
try.” There is now a relatively long history of examining the nature of
consciousness in animals and the 20th Century
“controversy” over discussions of animal or even human minds, and the
soft taboo about studying the mind directly, has justifiably receded
into the distance.
- Creative correlates of consciousness
Creative output is another source of data for assessing the presence of
consciousness. If for whatever reason we can’t examine neural or
behavioral correlates of consciousness, we may be able to examine the
creative products of consciousness (CCC) for clues. Additionally, in any
circumstance where the presence of consciousness is in doubt it will be
beneficial to use as many tests of consciousness as are available,
including NCC, BCC and CCC.
For example, when we examine ancient architectural structures such as
Stonehenge or other megalithic structures, or
cave
paintings in Europe that have been judged to be as much as 65,000 years
old, are we reasonable in judging the creators of these items to be
conscious in ways similar to our own? Most of us would say: obviously,
yes. We know from experience that it would take high intelligence and
consciousness to produce such items today, so we reasonably infer that
our ancient ancestors had similar levels of consciousness.
What if we find obviously unnatural artifacts on Mars or other bodies in
our solar system? Do we reasonably infer that whatever entities created
such artifacts were conscious? It will depend on the artifacts in
question, but if we were to find anything remotely similar to human
dwellings or machinery on other planets, but which was clearly not human
in origin, most of us would reasonably infer that the creators of these
artifacts were also conscious.
Closer to home, Artificial Intelligence (AI) has produced increasingly
impressive
art, with one piece fetching
over
$400,000 at a 2018 art auction. At what point do reasonable people
conclude that sophisticated art creation suggests the presence of some
kind of consciousness? To answer this question empirically, we could
conduct a kind of “artistic
Turing Test” and ask
study participants to consider various works of art and say which ones
they conclude must have been created by a human. And if AI artwork
consistently fools people into thinking it was made by a human, is that
good evidence to conclude that the AI is at least in some ways
conscious?
I reserve judgment on this issue for now, but as a general observation I
agree with Koch 2019 that any AI instantiated in a von Neumann
“feedforward” type of computer, which has no feedback processes
built-in, as are prevalent in brains, it is highly unlikely that any
complex consciousness will be present in any AI that is otherwise
impressive in its achievements.
- How do we develop a reliable psychometer?
There is not yet any widely-accepted or reliable psychometer, and
probably never will be a single device that reliably measures the
capacity for consciousness in all circumstances. But various researchers
have suggested ideas, including Dehaene, Changeux, Tononi, Koch,
Cassarotto, and others.
Demertzi admonishes when assessing the presence of consciousness:
“Finding reliable markers indicating the presence or absence of
consciousness represents an outstanding open problem for science.”
(Demertzi et al. 2019). Demertzi and colleagues, and various other
researchers have been working to identify reliable markers, but this is
still, as we have seen, a nascent field.
Dehaene 2014 states the problem clearly (p. 211): “[C]ould any
brain image ever prove or disprove the existence of a mind?” He answers
this question in the affirmative, with various discussions about the
neural correlates of consciousness and “signatures of consciousness”
(what he considers to be the necessary and sufficient correlates of
consciousness), but also recognizes that (p. 214) “no single test will
ever prove, once and for all, whether consciousness is present.” He
instead recommends a battery of tests be developed to give more
confidence about the presence of consciousness in various contexts. His
work is focused on human subjects but he also discusses animal
consciousness in his book. The present paper is focused on a theoretical
framework for measuring the capacity for consciousness in any physical
structures.
Neuroscientist Giulio Tononi and his colleague Christof Koch, as
discussed above, focus on measuring MICE and “integrated information”
as a proxy for measuring the capacity for consciousness. This theory
suggests that anything that integrates at least one bit of information
has at least a tiny amount of consciousness. A light diode, for example,
contains one bit of information and thus has the most rudimentary type
of consciousness. With just two possible states, on or off, however,
it’s a rather uninteresting kind of consciousness.
Koch 2019, accepting IIT as the best working theory of consciousness
extant, raises the possibility of panpsychism expressly, worth quoting
at length (Koch 2019 p. 160):
To the extent that I’m discussing the mental with respect to single-cell
organisms let alone atoms, I have entered the realm of pure speculation,
something I have been trained all my life as a scientist to avoid. Yet
three considerations prompt me to cast caution to the wind.
First, these ideas are straightforward extensions of IIT – constructed
to explain human-level consciousness—to vastly different aspects of
physical reality. This is one of the hallmarks of a powerful scientific
theory—predicting phenomena by extrapolating to conditions far from
the theory’s original remit. There are many precedents—that the
passage of time depends on how fast you travel, that spacetime can break
down at singularities known as black holes, that people, butterflies,
vegetables, and the bacteria in your gut use the same mechanism to store
and copy their genetic information, and so on.
Second, I admire the elegance and beauty of this prediction. The mental
does not appear abruptly out of the physical. As Leibniz expressed it,natura no facit saltus , or nature does not make sudden leaps
(Leibniz was, after all, the co-inventor of infinitesimal calculs). The
absence of discontinuities is also a bedrock element of Darwinian
thought.
Intrinsic causal power does away with the challenge of how mind emerges
from matter. IIT stipulates that it is there all along.
In my work on General Resonance Theory (GRT) (Hunt 2011, Schooler, Hunt
and Schooler 2011, Hunt 2014a, Hunt 2014b Hunt 2019), my collaborators
and I share this “panpsychist” foundation with IIT and other
panpsychist theories of consciousness. Indeed, Koch wrote the foreword
to Hunt 2014b. I accept as a working hypothesis that any physical system
has some associated consciousness, however small it may be in the vast
majority of cases.
Rather than integrated information as the key measure of consciousness,
however, GRT focuses on resonance and synchronization and the degree to
which parts of a whole resonate at the same or similar frequencies.
Resonance in the case of the human brain generally means shared electric
field oscillation rates, such as gamma band synchrony (40-120 Hertz) as
one example. (Hunt and Schooler 2019 also speculates that quantum
resonance may also play a role in human and other mammalian
consciousness, in an interplay with electric field resonance).
A psychometer developed pursuant to GRT would, insofar as it focuses on
neural correlates of consciousness, look at the degree of
shared resonance of various types (Hunt and Schooler 2019, Hunt 2019b),
and resulting information/energy flows, as the measure of consciousness. Humans and other mammals enjoy a particularly rich kind of consciousness, because there are many levels of pervasive shared synchronization throughout the brain, nervous system and body (Hunt 2019c). [Fn 2]
Footnote 2. Information is generally
defined as a subjective aspect of the physical world, whereas energy
is an objective aspect; but in this context I am using these terms
interchangeably because I define information as aspects of energy that
we can measure. As such, all physical dynamics consist of nothing more
than energy flows, but those energy flows that we can measure may be
labeled “information” and quantified under established information
theoretic concepts. I will, however, generally refer to
“information/energy flows” simply as “information flows” from now
on in this paper, for simplicity’s sake.
In our framework more generally, we propose a “weight of the evidence”
approach to assessing the presence and nature of consciousness in any
particular object of study. We pose a number of “questions,” in all
areas of the MCC as described above, to the object of study and it
answers in whatever ways it can. Questions can be verbal in nature, or
physical probes, or any kind of interaction between the tester and the
testee. Based on whatever responses are received, we then make the same
kinds of reasonable inferences about the presence and nature of
consciousness that we do every day, implicitly, when it comes to other
humans or animals.
This question and answer process is meant to be truly general and may
apply to any candidate for consciousness, whether it is a human, animal,
plant, bacterium, AI or any physical object.
The logical chain of this framework is straightforward: I know I’m
conscious; I assume other humans are conscious because they act in
various ways like me and do many intelligent things; I engage in similar
reasonable inferences when assessing whether various animals are
conscious and to what degree; we can use the same process of reasonable
inference all the way down the chain of physical complexity. Figure 2
summarizes this approach.
Figure 2. Summary of the suggested approach for assessing the
presence and nature of consciousness in any physical structures.