Metacognitive Skill In Emotion Regulation Stanford

TITLE: Metacognitive Skill in Emotion Regulation | Stanford CHANNEL: Meta-Think DATE: 2026-04-14 URL: https://www.youtube.com/watch?v=lVatZFt0Fu8 ---TRANSCRIPT--- All right, why don’t we go ahead and get started? Yeah, okay. So, Brendan, the floor is entirely yours. Sure. Hello, so I’m Brendan Conway Smith. I’m happy to be sharing my research with you on metacognitive skill learning. My background is in psychology and cognitive science, so there’s strong a cog sci angle in this talk, but my interest in this field got started in emotion regulation, which I think can really is crucial to unleash human potential. So, throughout my PhD research, the question that’s been motivating my research has been what are the cognitive mechanisms underlying metacognitive skill learning. And I’ll be trying not to overload you with information, and I’ll be instead giving you more of a research story about how I’ve been developing a theoretical framework through which we can understand metacognition as a kind of skill, appealing to philosophy and psychology literature to fill this out, and then apply it to the case of emotion regulation. And in particular, I’ll be describing a novel metacognitive skill learning theory that I developed, which explains how metacognition automatizes, how this helps to explain metacognitive learning in many different subdomains, and an extension of this is the hypothesis and predictions that my theory developed, and I’ll be exploring as potential postdoctoral research.

Okay, so we’ll start with part one, metacognitive skill. What does this mean? So, through history, people have been very interested in what makes expert performance. Expert performance has been a great value, whether you’re a hunter-gatherer or a modern doctor. And people have noticed that there really is a world of difference between a novice and an expert in all domains, and it seems to exhibit different and distinct characteristics shared within all domains. So, for example, the perceptual motor domain, maybe you remember what it was like to learn to drive for the first time. Your performance was slow, effortful, error-prone, at least mine was, but something happens as you practice. What happens is performance gets faster, more effective, less effortful, less relying on working memory, and more flexible. You’re able to adapt it to different situations. So, we see this in motor skill, and researchers have also shown that it exists in the cognitive domain, whether you’re developing math proficiency or chess. Initially, people start off with slow understanding of the rules and the strategies, how the pieces move. It’s very error-prone, it’s very clunky, but over time, again, your performance becomes faster, more effective, less effortful, more flexible, and more intuitive. You just sort of get a feeling for how it works, sort of like with driving. And in the field that I’m particularly interested in, metacognition, we see a lot of the same characteristics. For example, in emotion regulation, people implementing a tactic for the first time might be very slow, very error-prone, might be very effortful to apply some emotion regulation tactic, but over time, as they practice, performance gets more effective, they’re faster, becomes less effortful, and people are more flexibly able to adapt their emotion regulation strategies to changing dynamics in different environments. And this is really the ideal, and the question is how do people develop expertise in emotion regulation? And this is a meta-level skill that I will be referring to by appealing to this famous framework by Nelson and Narens, which distinguishes between two levels of cognition, the object level and the meta level.

So, in the object level, you have cognitive processes that direct performance towards external tasks. So, we talked about driving or tennis, for example, or cognitive skills such as chess or math, the performance, the objective cognitive processes are external, they’re world-oriented. They’re directed outward, and this is distinct from meta-level processes, where the information flow is towards this first-order cognitive process, where monitoring and control processes are directed towards internal targets, such as attention, emotion regulation, memory, and this area, the object level of cognitive processes and performance has been well studied over decades. They have common features, but it is this meta level that hasn’t been studied quite so much, and this is what interests me.

And this is the problem that metacognition itself has been difficult to understand, and metacognitive skill much more so. The research supports that people improve in all the different subdomains, such as attentional control, emotion regulation, such as memory, but we don’t know how. The question is what are the cognitive mechanisms involved in metacognitive skill? Previously, there’s been no formal theory on these mechanisms, which limits its application in education, therapy, research, many different areas. And an ideal framework, an ideal theoretical framework would help to integrate the empirical data, inform application, would help to clarify how improvements in monitoring and control processes occur in all the different subdomains. This has been one area of research. Actually, it’s been the main focus, I should say, of my master’s and PhD research, where I published several papers on my theory of metacognitive skill acquisition, and I believe these psychological insights can help to inform emotion regulation research and its application.

So, okay. Metacognition, as I always say, is a highfalutin term for something we all do and we perform all the time. So, we all know what it’s like to steer our attention towards some task that’s important, or try to upregulate some emotion or motivation in order to, you know, do the tasks we find challenging or or important. So, metacognition is really this umbrella term for the ability for cognition to monitor and control its own states. It’s been commonly referred to as thinking about thinking, which is partly true, but not entirely true, because it involves emotion, motivation, monitoring and controlling feelings, these types of things. It’s been with different neurological areas, such as the rostral lateral prefrontal cortex, but that will be the the subject of my talk. I’ll be referring more to these cognitive processes, which involve monitoring, the perceiving of cognitive states, then control, the regulating of these cognitive states.

And researchers have found it useful to apply this dual process lens, which has been popularized by Danny Kahneman in the sort of fast and slow type process, but rather applied to metacognition. Research shows that there are these two types of processes, where you can have monitoring and control driven by affective signals, procedural metacognition, where we have feelings of knowing, it’s fast, it’s automatic. And type two, we have metacognitive processes that are more concept-driven. They’re driven by explicit strategies, they’re slow, they’re effortful, they’re deliberate. And I just want to ground these in your own personal experience. So, type one, procedural metacognition, which is more, as I said, affect-driven, driven by feelings, driven by implicit signals. It’s fast, it’s automatic. For example, this famous tip-of-the-tongue phenomena, where we have a feeling that we perhaps know the name of an actor in a movie, we just can’t recall the name itself. Metacognition was coined by John Flavell, who was a Stanford professor in his work on metamemory, and he just discovered in people’s experience of of recall during these tests, these memory tests, that people had a pretty good ability to predict whether they would remember something. They had feelings of knowing that reliably tracked their success at recall. This can also refer to interoceptive awareness, metacognitive sensitivity, but importantly, this is not the type of metacognition that can be trained as a skill. This is more an architectural given that we just sort of get for free, as they say.

Type two metacognition is largely the area that we can train, where it refers to propositionally structured explicit concepts or strategies that are slow, they’re more conscious, but largely refers to working memory being largely taken up by these complex propositionally structured facts and instructions. So, we can have facts that refer to our own cognition, such as in the case of memory. We can say, I can’t remember that person’s name, the actor’s name from that movie, or an emotion, you may say, I feel frustrated. These are explicit, verbally expressible concepts, but they don’t drive cognition. Instructions are largely the information type that are causal control models that refer to directing one’s own cognitive processes towards certain goals. These are perhaps attentional or emotion goals or memory goals. So, a metamemory strategy might be, if you can’t remember somebody’s name, sound out the letters of the alphabet as you go through the alphabet, it may trigger prompt memory itself. Or in emotion regulation, you may have this instruction for reappraisal, where you reorient your goal. If you’re having some sort of argument with a loved one, you may think that I want to protect the relationship, so restrain my anger, maybe feel more compassion or empathetic. So, these are propositionally structured concepts and instructions that can help to drive control processes, and these are the types that can get trained up over time.

So, metacognitive skill, it is the extent to which we can monitor and control our own processes. Metacognitive skill is recognized by global organizations as a vital 21st century competency. Research shows that children who cannot regulate emotions are less able to learn, develop healthy relationships, cease bad habits, which is more important than ever. And interestingly, it can even outperform IQ as a predictor of learning and performance outcomes.

And these control processes are largely what we’ll be talking about in terms of skill, and there are many different ways we can monitor and control our own cognitive processes. We can engage in learning, deliberately attentional processes, reasoning. Interestingly, science itself can be considered one large metacognitive enterprise. And so, my interest was, what are the common mechanisms? What are the common features and factors underlying all these different ways that cognition can direct its own processes? Now, my primary focus and my whole motivation for getting into this field were skills related to emotion regulation. Uh of course, there is reappraisal, where you’re deliberately changing, reinterpreting, reframing the meaning of some negative emotion in order to lessen its valence. There’s equanimity, which I find particularly interesting, where you’re deliberately perceiving the moment-to-moment changes or the impermanence of an emotion to create this helpful disengagement that decreases meta emotions. And there are attentional skills you can learn that allow you to detach from unhelpful, maladaptive thinking patterns such as rumination and worry.

And this is consistent with literature showing that mental health is increasingly being framed in terms of metacognitive skillfulness. With disorders viewed as deficits in this ability, which can be ameliorated by developing greater skillfulness in attaining emotion regulation goals. And we see this largely in therapeutic techniques, particularly in cognitive behavior therapy training, where people regularly practice metacognitive skills such as emotion regulation skills such as cognitive restructuring, mindfulness, all types of different skills.

And the question that I’ve been interested in is, what are the characteristics of emotion regulation skill as as metacognitive features? So, the characteristics of skill are fascinating area, where there have been decades of research on skill learning, motor skill, cognitive skill. This research has been done in psychology, in philosophy, cognitive science, neuroscience. And this is this very interesting ability for the human brain to improve at almost anything. And we talked about these shared characteristics of expert or elite level performance, where it’s fast, it’s effective, it’s it’s largely effortless, it’s flexible.

And so, skill theorists refer to this idea of control, where elite performance refers to a high degree of control that an agent possesses over some activity in some domain, often requiring years of practice. So, it’s this ability for an agent to align their actions with a goal-directed plan, explicit plan, flexibly, reliably, consistently. And so, a recent publication of mine clarified these characteristics within metacognitive skill.

And so, importantly, a skill theorist must explain what the intelligence of skilled action and skilled control consists of, how it gets trained up, how it gets more effective, the underlying forms of information, the information processes, the representation types. So, we already talked about perceptual motor skill, driving, tennis, cognitive skill, math, chess, metacognitive skill such as reasoning, emotion regulation, memory.

And there’s been a lot of research in the bottom two, but not so much on the top. So, researchers have largely focused on motor and cognitive skill and have discovered common features that underlie these processes, which distinguish an expert from a novice, which are involved shared cognitive, neural, computational components that experts use to achieve extraordinary performance. These common factors that they share involve goal structures, action types, knowledge types, the fact that they can automatize, they can become more flexible. And so, I’ve been largely trying to bridge the literature between skill and metacognition to help clarify and make this field more clear, because it has been bedeviling researchers for a long time, and it is a tricky domain.

So, I want to start with goal structure. Well, I think we’ll all clarify some of the questions that have been asked so far. It is this question as to what distinguishes cognitive control from metacognitive control. It seems to be a bit of a fuzzy boundary. And metacognitive control at the expert level requires the ability to flexibly and reliably align our deliberate mental actions with a plan directed at metacognitive goals. And so, cognitive control regulates cognitive processes instrumentally, making use of them as a means to attaining some externally directed goals such as playing chess or finding that coffee shop. So, they’re externally directed goals. And when it comes to metacognitive control, it refers more to processes that target mental processes or states directly as goals or ends in themselves. So, these are explicitly structured goal states, and different subcategories of metacognition can be distinguished by their different metacognitive goals. So, you can have the metacognitive goal of achieving truth or better attention or affect or better memory. And so, in terms of emotion regulation, it requires the controlling of mental actions to achieve some affective goals such as reducing depression or anxiety, where you apply some conceptual tactic such as reappraising stress as a positive growth signal to reduce the negative balanced emotion.

And this helps to distinguish between these two levels, where you have object-level cognitive goals that explicitly refer to world-oriented ends, towards which control processes are directed. And then you have metacognitive goals, which are explicit goals referring to internally oriented ends, towards which control processes are directed.

So, we can have attentional goals, greater flexibility or the greater all-inclusiveness in mindfulness practices, uh reappraisal, which refers to as Lang refers to the second-order attitudinal goals, where you’re specifying the stance that one takes towards one’s own mental states. And I agree with Dr. Gross, who a few weeks ago referred to equanimity or the deliberate perception of affective change, this moment-to-moment impermanence of experience as a type of reappraisal, as you are engaging in some second-order attitudinal goal of changing your stance towards your own mental states. So, this can help to clarify uh the difference between cognitive goals, object-level goals versus metacognitive or meta-level goals.

Again, there are many different possible metacognitive goals. And in emotion regulation, Prime, Bacon, and Gross have referred to these three different goal types, which involve upregulating, downregulating, neutralizing, and transforming goal states. Emotion regulation goals can be implicit such as a vague desire, which can be hard to go about trying to achieve, or an explicit goal, a thought referring to, for example, decreasing the negative balance of some emotion. And they are uh pervasive throughout the process model of emotion regulation, throughout these different families of emotion regulation strategies, where you’re trying to change perhaps some situation, modify it, try to deploy some attentional process, response modulation. But I believe it’s this area of cognitive change, it’s this stage in the process model that I think would benefit from being conceptualized as a form of metacognitive control. I believe it already has been in many respects. And in terms of the extended process model, uh metacognitive goals, emotion regulation goals refer to this area of identification, where one is specifying a goal of how one wants to feel, how one feels, how one is attempting to feel.

And importantly, this also necessitates uh subgoals and skill. As we all know, it’s hard to just will ourselves to achieve goals. Anything above a certain level of complexity requires subgoals to achieve. Work by Craft and colleagues have argued that in fact, all psychotherapy can be seen as emotion regulation goals that they attempt to achieve through patients’ hierarchically structured affective goal states. Subgoals are organized according to their conduciveness. So, in, for example, perceptual motor skill in driving, your goal is to, you know, arrive. And subgoals involve steering, braking, gas, which require further subgoals such as observing signs, watching for pedestrians. And in the cognitive domain, when you’re playing chess, the goal is to checkmate the opponent, which will involve subgoals such as opening moves, attacking the opponent, which involve further subgoals such as avoiding traps, controlling the center of the board. And in emotion regulation such as reducing anxiety, you have subgoals, which involve maybe choosing a strategy, selecting a tactic, which necessitates further subgoals such as focus, monitoring effectiveness, resisting intrusive thoughts. And this relates to these areas of selection and implementation, where you have the tactic being appropriate for the goal and the context, which is then implemented.

I wanted to also refer to this work by Craft and colleagues on meta emotion regulation, which refers to subgoals for more helpfully achieving emotion regulation goals. So, meta emotional regulation in this term refers to the processes that influence how you go about the actual emotion regulation tactics. So, they change your emotion regulation behavior itself, instead of the actual emotion directly. For example, during a tense meeting, a manager might reframe criticism as useful feedback to stay calm and constructive. And in meta emotion, this would be a subgoal that manages how you will implement the tactic to regulate that emotion during the meeting. So, the manager might say, before the meeting, they’d set a rule, “Okay, you know, if this discussion gets tense, I will pause, breathe, and try to reinterpret the comment as problem-solving rather than personal criticism.” So, it’s a way of helping to motivate and direct people to be better at actually implementing emotion regulation strategies.

So, clustered actions is another really important aspect of skill, where expertise in any area usually involves being skilled at more than just one action. It requires a cluster of action types. These are deployed hierarchically. For example, basketball, you may have footwork skills that involve dribbling to eventual layup to, you know, win the game. In math, you learn basic math skills in order to uh solve more complex equations and emotion regulation, you have perhaps clarifying the goal, clearing the mind, focusing before attempting some reappraisal. You also have goals that can be applied simultaneously. Um this involves polyregulation or for example, in tennis, an overhand serve might be uh combined with the appropriate body position in chess. You may be attacking an opponent while trying to avoid a trap. In emotion regulation, you may be simultaneously resisting mind wandering while attempting some compassion exercise while also trying to uh resist intrusive thoughts, these types of things.

Lastly, before the break, I referred to different knowledge types. This is important that the skill literature distinguishes between these two types of expert knowledge, uh declarative knowledge, which is correlated with the temporal medial lobe and working memory, where they are, again, propositionally formatted, explicit goals, rules, tactics, but declarative knowledge doesn’t necessarily execute the action. They can refer to the right action types to take, but they don’t execute the action. This requires procedural knowledge, which is correlated with the basal ganglia. These are the implicit representations that actually execute the action and the control to achieve goals. So, these are the the state-changing representations of the system that actually execute the actions and the instructions themselves. So, whether it’s tennis, chess, or emotion regulation, these require agents to use declarative knowledge to learn which tactics achieve the goals, and then procedural knowledge to execute the actions themselves. In fact, metacognitive has been referred to as domains of procedural knowledge that are trained up specifically, and I believe emotion regulation tactics can align with this. And so, interestingly, as slow declarative knowledge is repeatedly practiced, it’s then replaced by fast, automatic procedural knowledge. And so, these actions then proceduralize, then they can become these automatic metacognitive, physical, and cognitive control habits that run offline outside of working memory.

So, skill theories have referred to this concept of proceduralization, which is exactly what we were just talking about, which explains how metacognition automatizes. So, for example, early in learning, you have these slow, explicitly driven processes that rely on conscious, deliberate, effortful, very error-prone performance. Then over time, they become more explicit. So, these slow processes are replaced by procedural operations, which enable faster, more efficient implementation of tactics with minimal cognitive effort. So, this theory had not been applied to metacognitive skill or emotion regulation, which I found fascinating. So, in the first stage, we have these type two explicit processes or instructions akin to be triggered, so you have task knowledge that’s acted out and triggering procedural knowledge to carry out performance. It’s immediately working memory heavy. In the intermediate stage, it builds up procedural knowledge that then is repeatedly practiced, and you have performance speed go up, errors go down. In the automatic stage, you have actions that rely largely on procedural knowledge. It becomes more automatic, faster, more accurate. It gets trained up in these domains. It operates outside of working memory with minimal cognitive demands, and it follows this power law speedup. And it’s the process where type two processes become these implicit type one operations.

And so, the phenomena that I’ve helped explain this with that I published refer to the attentional training technique, where we can automatize attentional control processes that can flexibly shift from maladaptive patterns of thinking and feeling. So, from rumination and worry, you can you’re more easy to recognize and detract from these rather than just going about them automatically. The metacognitive threshold, which involves how people become more emotionally aware through mindfulness training, where you refine procedural knowledge that allows for the detection of subtler affective signals, which allows for more skillful regulation. And detached mindfulness, which builds on this, which is how equanimity, this perception of of affective impermanence, disengages meta-emotions because meta-emotions can be largely these implicit processes, and they can be replaced by more skillful procedural knowledge, where it’s trained not to react, to engage in this sort of decentering.

And in emotion regulation, I believe that as procedural knowledge builds up, you’d have emotion regulation goals achieved more efficiently, more effectively, more flexibly. You’d see the characteristics of skill be expressed more easily. And this helps to explain a couple things, such as this hallmark of skill, flexibility, allows us to regulate our emotion regulation efforts in situational demands. As proceduralization builds up, it offloads tasks from working memory. This frees up working memory for higher-level control. This is also called cognitive reinvestment. Whereas working memory is freed up, it can be directed towards planning, it can be redirected towards error monitoring or adapting to novel or changing situations. So, for example, if you’re in a chaotic office workspace, a lot of dynamic situations going on there, and having more working memory in these automatized emotion regulation processes can be very helpful. And so, working memory allows us to better identify, select, and implement emotion regulation strategies, and this allows us to helpfully understand how all these skill areas are combined.

Since we’re sort of running out of time, I want to respect your time for these meetings, and I just want to really quickly talk about just power law practice that has been shown to be a general feature of skill learning. Whereas automatization builds up, we have these reaction times that follow power law function, where you have this steep initial improvement gradually leveling off. Uh you see this linear plot in this log-log plot, and it struck me that uh across domains, we see this power law function of implementation, and this allowed me to sort of predict that metacognition should follow this display in power law on learning results, where you should see the same rapid early gains followed by a slower progress as it approaches a performance ceiling. And this allowed me to look in the literature, and I did see a power law in these subdomains of metacognition, attentional control, and metamemory learning. You see this power law function of timing speedup. And so, one candidate postdoctoral project would empirically test this theory and its predictions to see whether emotion regulation displays these properties.

Uh I’m I’m just going to zoom through this, but the three markers would be implementation speed, cognitive effort being reduced, of course, increased effectiveness of these of these tactics, which allows for these testable hypothesis that participants who practice, for example, reappraisal would regulate emotions more quickly with less effort, with trajectories consistent with power law functions. Uh I won’t go through this study design quite yet. This is still quite general, but I want to get to something I find really interesting is this polyregulation, which proceduralization and emotion regulation helped us to understand as the deployment of many skills at one time can be helped to explain through this hierarchical deployment of skills that can be helped through this process of proceduralization, where you have these habits of mind that are beneficial, uh initially expressed explicitly. When they’re practiced, they become offloaded from working memory, they can become mutually supportive, so you can have many of them running at once. Uh reappraisal, attention, mindfulness to allow for more effective emotion regulation goal attainment. And um Yeah, so I just wanted to zoom through that since I don’t want to go over time and I get to any questions, but these are some theoretical and empirical extensions of this metacognitive skill theory that involve uh temporal dynamics, help to provide some functional clarity that I hope might lead to some refined clinical interventions.

So, uh more questions, please. Great. Thank you, Brandon. That was really Brandon, we’re we’re actually at time. Uh so, we don’t have time for broader conversation, but I’ll encourage everyone in Zoom land and also here in the room to email you or follow up for a broader conversation and we’ll continue. Uh but thank you. Thank you. Thank you. Thank you very much.