Know Thyself: The Science of Self-Awareness - Stephen M Fleming 2021
Emulating Socrates
The Power of Reflection
I am not yet able, as the Delphic inscription has it, to know myself; so it seems to me ridiculous, when I do not yet know that, to investigate irrelevant things.
—PLATO, Phaedrus
We have come full circle. We started our journey through the science of self-awareness with the ancient Athenian call to know ourselves. We now know that this is not an empty platitude. Instead, the human brain is set up to do exactly that. We track uncertainty, monitor our actions, and continually update a model of our minds at work—allowing us to know when our memory or vision might be failing or to encode knowledge about our skills, abilities, and personalities. By understanding how self-awareness works, we have seen how to use it better in situations ranging from the boardroom to the courtroom and begin to extract insights into the machinery and computations that made the human brain self-aware—insights that might inform how we build and interact with AI devices.
In some ways, though, we have ducked Socrates’s challenge to discover ways to know ourselves better. It is still early days for research focused on enhancing or improving self-awareness, but there have been some pioneering attempts to develop both high- and low-tech solutions to this challenge. The good news is that, as we saw in Part I of this book, metacognition is not cast in stone but can be shaped and molded by training and experience. This was not always believed to be the case. For much of the twentieth century, the received wisdom among both scientists and the public was that once the human brain had grown up, its circuitry became relatively fixed.
We now know that experience and extensive practice can alter our brain structure well into adulthood. London taxi drivers famously have a larger posterior hippocampus, a region presumably important for storing and retrieving “The Knowledge,” the body of information that they are required to learn about London’s streets. Skilled musicians show greater gray matter volume in their auditory cortex—again, presumably due to the specialization of this cortical region for processing sounds. There is also evidence that practice can directly lead to changes in brain structure. Learning to juggle leads to increased gray matter volume and white matter in regions of the parietal cortex involved in processing visual motion. In studies of mice, these kinds of volume changes have been linked to the expression of proteins associated with axonal growth, suggesting that brain-volume increases are the observable consequence of extra connections being formed in the brain as a result of practice and training. Extensive research on animals tells us that new learning is accompanied by changes in synaptic weights: the fine-grain molecular structure that allows one neuron to communicate with another.1
In other words, not only can adults learn new tricks, but we do not usually have any choice in the matter. The architecture of our brains is being subtly buffeted and updated by everything we do, whether we like it or not. The implication is that, just like other brain functions, metacognition is not fixed or immutable.
To get a sense of what might be possible here, we can take a closer look at experiments that have attempted to modulate or boost metacognition in different ways. In one experiment involving elderly volunteers, researchers at Trinity College Dublin explored whether passing a weak electrical current through the PFC—a technique known as transcranial direct current stimulation (tDCS)—could heighten metacognitive thinking. Subjects were asked to perform a difficult color-detection task under time pressure and to provide a judgment of when they thought they had made an error. In comparison to a sham stimulation condition in which no current was applied, the tDCS increased the volunteers’ awareness of when they had made an error, without altering overall performance on the task. We still have little idea of how tDCS works, but it’s possible that the weak currents temporarily excite neurons and place the PFC in a state that improves metacognition.2
In studies using a similar task, the drug methylphenidate (Ritalin), which increases the concentration of dopamine and noradrenaline, is also able to enhance people’s awareness of their errors. My colleagues at UCL have taken this a step further by showing that giving beta-blockers—typically prescribed to reduce blood pressure by blocking noradrenaline function—provides a significant boost in metacognitive sensitivity on a perceptual discrimination task. This was not the case for another group of subjects who instead took a drug designed to block dopamine function, amisulpride. Data on the pharmacology of metacognition is rare, but so far it suggests that boosting systemic levels of dopamine and inhibiting noradrenaline (which, incidentally, is also released in times of stress) may have overall benefits for self-awareness.3
It may even be possible to train people to directly alter brain circuits that track confidence in their decisions. In a study published in 2016, researchers at Japan’s Advanced Telecommunications Research Institute demonstrated they could train a machine learning algorithm to classify people’s confidence levels from the activity of the PFC while they performed a simple task. The researchers then asked people to practice activating those same patterns on their own. If they were able to activate the patterns, a circle on the computer screen got bigger; if they inadvertently began deactivating the patterns, the circle became smaller. There was no need for the subjects to know what they were doing—all they had to do was make the circle bigger, and the algorithm took care of the rest. After two days of training, the participants’ metacognition was measured again. In the group that was trained to boost “high confidence” brain patterns, their confidence was increased, whereas in subjects trained to boost “low confidence” brain patterns, confidence was decreased. These represent small changes in metacognitive bias, and because they experienced bigger confidence boosts on incorrect trials, the high-confidence group members in fact suffered in their metacognitive sensitivity. But these findings are a proof of concept that more targeted boosts to self-awareness may be possible.4
Many of us might be reluctant to receive brain stimulation or take drugs to boost metacognition. But we might be willing to invest some time in simply practicing being more self-aware. With this goal in mind, my lab has been working on developing training protocols that provide feedback not only on people’s performance, but also on their metacognition. We asked people to practice simple perceptual judgments (about which of two images was slightly brighter) for around twenty minutes per day. In one group of subjects, we gave people feedback about their metacognition—whether their judgments of confidence were accurate or inaccurate. A control group of subjects received feedback about their perceptual judgments—whether they chose the correct image. We found that people in the metacognitive feedback group had heightened metacognitive sensitivity after two weeks of regular training.5
This promise of being able to improve and reshape our self-awareness depends on whether this kind of training is likely to have benefits that go beyond the laboratory. In our training experiments, improvements in metacognition were also seen in a memory task that was not part of the training. In other words, having learned to boost metacognition on one task, people were naturally able to transfer their newfound skills to better reflect on their performance on a different task. This suggests that the skills we are training may be broad and domain-general. Consistent with a role of metacognition in promoting more open-minded and considered decision-making, we’ve also found that these self-awareness boosts lead to a systematic improvement in recognizing when you might be wrong and when more information might be useful. There is also some evidence that these boosts in metacognition, installed by practicing a simple game, might be sufficient to promote more open-minded decision-making about contested issues such as climate change.6
It is worth sounding a note of caution here. The improvements in metacognition we see in the lab are small, and—aside from potentially beneficial effects on decision-making—are unlikely to have a noticeable impact in our daily lives. However, we should be mindful that more substantial boosts in self-awareness may not be an unalloyed good. In cases such as dementia, a well-intentioned attempt to boost insight and provide patients with awareness of potential memory failures might create unwanted anxiety and depression. Even in the otherwise healthy brain, we have seen that creating more realistic self-appraisals may have emotional downsides, and small doses of overconfidence may be useful for cultivating self-efficacy. Just as we would want to carefully evaluate the side effects of any new drug, we should also be mindful of the potential side effects of boosting self-awareness.7
But I hope that by now, I have convinced you that the benefits of improving self-awareness can greatly outweigh the negatives, especially when it is finely tuned. We have seen that self-awareness provides a spark for a whole gamut of human culture—from cave art to philosophy to literary novels—and helps us live in harmony with others, especially when they hold views that oppose our own. Set against this background, Socrates’s suggestion that we should spend time boosting our self-awareness seems more relevant than ever.
More Than Conscious
If it were possible to systematically improve or alter self-awareness, what would that feel like? The answer to that question depends on how we think about the relationship between metacognition and awareness of the world around us. To see this, imagine that I put you in a dark room and asked you to watch for a faint light to appear on a computer screen. When the light appears, it presumably causes a difference in your visual experience. But it also leads to a difference in self-awareness: you think about whether you saw the stimulus or not, and this reflective judgment allows you to notice and communicate a change in consciousness. How should we think about this relationship?
There are two dominant camps in the field. The first approach is to suggest that self-awareness is just an optional veneer on regular conscious experience. This “first-order” view accepts that we might need metacognition to reflect on and report our experience, but argues that this does not mean it is involved in creating the experience itself. Instead, metacognition is an optional add-on, layered on top of consciousness and merely required for getting the information out into the outside world.
One argument in support of the first-order view is known as the overflow thesis, which goes something like this: The world around you appears rich and detailed. From my desk, I can pick out the changing shades of white on the wallpaper as the light from my Anglepoise lamp plays off the wall, the wood pattern of my desk, and the deep blue of a painting on the wall. I could go on, but it seems obvious that there is far more in this scene than I can talk about. My private inner world is poorly served by our ability to access and describe it—it “overflows” the capacity for report.8
Overflow seems intuitive, and it has been used to support the idea that our experience and our ability to comment on that experience are two different things. But we should be careful before accepting this conclusion. Data from overflow experiments is also consistent with us having only a vague impression of a whole set of objects, which is brought into clearer focus by the requirement for report. It’s fiendishly difficult to know for sure what people’s primary conscious experience is like without asking them to reflect on what they are experiencing.9
An alternative camp is known as the “higher-order” view. Higher-order theorists propose that being able to reflect on and generate thoughts about our mental states is exactly what makes us conscious in the first place. Without metacognitive awareness of our mental states, we might still be able to process information or react to stimuli, but there would be no consciousness to speak of.10
There are some key predictions that higher-order and first-order theorists disagree on. For instance, to the extent that the PFC is important for metacognition and higher-order thought, then if two experimental conditions are created that differ only in conscious awareness, you should detect a difference in the PFC under the higher-order view but not under the first-order view. There is initial data that supports this hypothesis. The problem, though, is that the technologies we currently have for studying the human brain (such as fMRI and MEG) are too coarse to provide a strong pronouncement either way. We do not yet have unfettered access to the fine-grain patterns of neural activity underpinning subtle changes in the conscious and metacognitive states of human subjects—and so the jury remains out.11
There is, however, at least circumstantial evidence for a connection between metacognition and consciousness. The studies of blindsight we encountered earlier suggest a lack of metacognitive sensitivity is a feature of information processing in the unconscious (blind) hemifield. Conversely, damage to the PFC does not only impair metacognition, but it can also have consequences for conscious experience.12
Another, more subtle implication is that if metacognition is inherent to conscious experience, it may be difficult for us to reflect on consciousness itself. We have already seen that we rely on metacognition to distinguish between reality and imagination, to realize that our perceptual experience may be in error, and to recognize that we might be fooled by illusions. We may look at a visual illusion such as the checkerboard we encountered in Part I, and say to ourselves, “I know the squares are the same shade of gray, but it still seems like they are different.” I can recognize that what I see and what I think I see are in conflict. Encoding these differences between perception and reality allows us to have an awareness of seeing—an important component of subjective experience. But it seems much more difficult to make consciousness itself visible in this way. To us, consciousness just is—it is transparent. This transparency may even be the root of the hard problem of consciousness: why we think that there is a mystery of consciousness at all.13
My personal view is that this debate will hinge on which version of consciousness we care about more. It certainly seems possible that primary consciousness could exist in a vacuum, without our ability to introspect about it—to have a raw experience of smelling coffee, or seeing the color red, without engaging in metacognition at all. However, I also suspect that the kind of consciousness we cherish—the kind that allows us to appreciate the smell of coffee and tell our friends about the sunset—involves meta-awareness, that is, a state of knowing that we are conscious. This higher level of awareness is not just an optional veneer or gloss on experience, but instead a base layer of what it means to be human.14
Consciously experiencing emotions may also involve this kind of higher-order awareness. The neuroscientist Joseph LeDoux suggests that the bodily reactions that often accompany emotional states—such as freezing in response to a loud sound or sweating in response to a snake or spider—may be distinct to the conscious, reflective experience of shock or fear. He points out that laboratory mice and humans show similar automatic reactions governed by circuits deep in the brain when faced with danger, but it is likely that only humans can think about and become aware of what it means to feel fear.15
An interesting test for exploring the meaning of these different varieties of conscious experience is the puzzling case of dreams. It certainly seems that we are conscious in our dreams. But at the time of dreaming, it is rare that we have any self-awareness of being in a dream. This renders the experience a bit frustrating and meaningless, at least until we wake up. Wouldn’t it be great if we could sometimes become conscious of our dreams?
In fact, some people do report experiencing self-awareness during dreaming, in states known as lucid dreams. Lucid dreams tend to occur when people are overly tired, though it may also be possible to train yourself to become lucid at will. The frequency of lucid dreams also varies substantially between people—some never experience them, others experience around one lucid dream per month, and yet others dream lucidly most nights. I have only experienced a lucid dream once, when I was sleeping on a boat moored in the Hamble River on the south coast after a few days’ sailing. I suddenly became aware of flying through the boat and being able to guide myself around at will. It was a bizarre and wonderful experience. It felt like being fully awake and conscious, but in another world.
It is possible to track whether someone has gone into a lucid state by asking them to make a telltale set of eye movements that can be monitored while the person is asleep. Remarkably, brain imaging studies have shown that activity in the frontopolar cortex and precuneus is increased during lucid dreams—two regions that we have seen are also implicated in metacognition. Even more remarkably, electrical stimulation of the PFC at a particular frequency is sufficient to increase people’s lucidity during their dreams.16
Research on lucid dreaming, then, is perfectly consistent with the notion that people are not meaningfully aware of their dreams under normal circumstances. As in the case of overflow, it remains possible that we have something like primary consciousness without self-awareness during dreams, but it would seem difficult or impossible to tell whether this is the case with the current measures we have available. Instead, the data shows us that when people become conscious of their dream—when they become lucid—they recruit the same brain networks that support waking metacognition.
Becoming Lucid
I find it appealing that boosts to self-awareness in our daily lives might feel like the experience of becoming lucid in a dream—we might begin to notice things that we have not noticed before in ourselves, in others, and in the world. This is in fact strikingly close to the kinds of changes in consciousness that expert meditators report experiencing after intensive retreats.
It might be no surprise, then, that another promising route toward boosting metacognition and self-awareness is by engaging in regular meditation. Mindfulness meditation is central to Buddhist practice and has been closely linked to the kind of higher-order, reflective awareness that is characteristic of metacognition. But the impact of meditation on scientific metrics of self-awareness has only recently been explored. The data is encouraging. A 2014 study led by psychologists Benjamin Baird and Jonathan Schooler at University of California, Santa Barbara, reported that engaging in two weeks of meditation training increased metacognitive sensitivity during a memory test. Other work has shown that meditation experts have greater metacognitive sensitivity compared to controls.17
This is a small but growing area of research, and much remains to be done to replicate and extend these findings. Meditation studies have proven controversial in neuroscience because it is often difficult to create agreed-upon definitions of what meditation training means. But I remain optimistic. One exciting idea is that because mindfulness meditation involves consistent self-focus and the ability to zero in on our mental states, it might also hone our ability for self-appraisal. And the power of periods of reflection for boosting performance need not be restricted to classical meditation. Researchers at Harvard Business School compared groups of trainee employees at the Indian IT company Wipro, assigning them to spend the last fifteen minutes of their day either reflecting on what they had learned (the reflection condition), explaining the main lessons to others (the sharing condition), or continuing their studies as normal (the control condition). Compared to the control condition, the employees in both the reflection and sharing conditions boosted their performance on a final exam by over 20 percent.18
We have already seen, though, that self-awareness and metacognition are fragile, and the scope of plasticity is asymmetric. In other words, the gains that we can achieve by boosting metacognition are nothing compared to the downsides of losing it. It is concerning, then, that self-awareness may be under increasing threat from modern life. In a culture of being efficient and productive, taking time to reflect on what we are doing seems to be an optional luxury. Smartphones and screens are dominating our waking moments and squeezing out times at which we otherwise might have stopped to think. As we have seen, excess stress and deterioration in mental health may also lead to the erosion of self-awareness. There is a danger that we will become trapped in a vicious cycle in which we are more focused on doing and less on reflecting, and become less and less aware of the benefits of high-quality metacognition. By understanding the factors that lead to self-awareness failure, we can take steps to stop this cycle before it begins.
In turn, a science of self-awareness helps us adopt a compassionate stance toward occasional failures of self-awareness in others. Just as for ourselves, the self-awareness of our friends and colleagues is continually in flux and buffeted by a range of hidden signals affecting whether they think they are right or wrong about a particular issue. People who hold opposite views to our own on polarizing issues ranging from politics to vaccines may appear “blind” to the evidence. But by recognizing that our confidence in our views is a construction, and prone to distortion, we can cultivate a more tolerant attitude toward others who may not agree with us.
Perhaps the most important element in protecting and cultivating self-awareness is something you are just about to finish doing: reading and thinking about the science of self-awareness. By pulling back the veil, even for a moment, on how metacognition works, we can gain a newfound respect for the fragility and power of the reflective mind. There is a beautiful symmetry to this. Simply by studying self-awareness, we may gain more of it. Two and a half thousand years after the Athenians wrote their words of advice on the Temple of Delphi, we are now in a better position than ever before to know ourselves.