Let's Measure Consciousness!

If you're driving, you're having a subjective experience of colors, sounds and vibrations. But does a self-driving car have a subjective experience? Does it feel like anything at all to be a self-driving car, or is it a zombie in the sense of having behavior without experience? This question of why and when matter is conscious is the essence of what philosopher David Chalmers has termed "the hard problem" of consciousness, and it's important not only in philosophy. For example, if you're an emergency room doctor, how can you determine whether an unresponsive patient is conscious in the sense of having a subjective experience? Patients with locked-in syndrome have functioning minds without being able to move or communicate. And what about a future robot intelligent enough to converse like a human?

A traditional answer to this problem is dualism -- that living entities differ from inanimate ones because they contain some non-physical element such as an "anima" or "soul". Support for dualism among scientists has gradually dwindled. To understand why, consider that your body is made of about 1029 quarks and electrons, which as far as we can tell move according to simple physical laws. Imagine a future technology able to track all your particles: if they were found to obey the laws of physics exactly, then your purported soul is having no effect on your particles, so your conscious mind and its ability to control your movements would have nothing to do with a soul. If your particles were instead found not to obey the known laws of physics because they were being pushed around by your soul, then we could treat the soul as just another physical entity able to exert forces on particles, and study what physical laws it obeys.

Let us therefore explore the other option, known as physicalism: that consciousness is a process that can occur in certain physical systems. Instead of starting with the hard problem, we can then start with the hard fact that some quark blobs are conscious and others aren't, which leads to the fascinating question of what makes the difference. I've long contended that consciousness is the way information feels when being processed in certain complex ways, but what types of information processing quality? Specifically, what mathematical equation much an information processing system satisfy to be conscious? Answering this question might allow future ER-physicians to have a consciousness detector, and would let future programmers control whether they built consciousness into their artificial intelligence systems.

Neuroscientist Giulio Tononi has proposed just such an equation, which forms the core of his Integrated Information Theory of consciousness (IIT). It says that information being processed is conscious if a mathematical quantity called Φ ("Phi") is large. Phi quantifies integration, the extent to which information is interconnected into a unified whole rather than split into disconnected parts. The theory has generated interest from the neuroscience community, but also controversy, including recent critique from Scott Aaronson.

I want to see the question of whether IIT is correct or not resolved by experimental tests. Unfortunately, Tononi's proposed measure of integration is too slow to compute in practice from state-of-the-art patient data, requiring longer than the age of our universe, let alone the lifetime of the patient. I've therefore worked hard over the last year in search of a faster way to compute integration, and I'm happy report that I've found one -- in fact, several. In a paper I just posted, I explore and classify existing and novel integration measures by various desirable properties, finding that although there at first seem to be a few hundred options, there are in fact only a handful if attractive ones. I was happy to discover that there's an approximation based on graph theory that let's you dramatically speed up the exact formulas, so that they can
be applied to real-world data from laboratory experiments without posing unreasonable computational demands. Let's try to shed more light on the fascinating questions and theories about consciousness my making experimental tests!