Emotion and the Dynamics of Representation Use
23 Mar 2026 - 30 Oct 2026
- Emotion and the Dynamics of Representation Use, a position paper I wrote for a workshop on emotions in AI in 1992. It's pretty half-assed, but the ideas are still interesting, at least to me.
- I'd absorbed critiques of representationalism from Agre and Chapman, but didn't accept their eliminitivism. So I wanted to come up with a way of thinking about representation that made sense.
- Also had been working with Tinbergen-esque theories of animal behavior, and was kind of in the mode of looking for simple cognitive mechanisms that were more complex than mere reflexes but made evolutionary sense.
- And was working under Marvin Minsky, trying to boil down aspects of his Society of Mind theory (which itself was rooted in Tinbergen and Freud) into something simple enough for me to understand and implement.
- I coined the term "enactive representation", perhaps influenced by The Embodied Mind. Or rather I thought I was making it up, but it was used by Jerome Bruner to mean pretty much the same thing.
Situated action theory [Agre and Chapman 1987] includes the idea of deictic representation, which combines this sort of “leaning-on- the-world” with the language of representation. It can be argued whether such constructs are really representational since they don’t involve mental state; they are more a way of talking about the relationship between the world and an actor’s dynamics.
The enactive view builds on the deictic view and attempts to integrate representations of things that aren’t actually available to the senses. The primary use of representation is imaginative, in that it is used to perceive what isn’t actually at hand. An enactive representer imaginatively re-enacts a past action in order to predict the result. Action becomes not only a matter of adaptive reaction to the world (as it is in situated action systems) but of reacting to imaginings about the world's future.
- Since then "enactive cognition" has become a minor academic industry.
- The paper suggests that the mechanisms I propose might underlie Freudian repression. This is a little embarrassing to read now, because Freud himself had very elaborately described models of how repression worked and what the dynamics of the machinery might be, which I guess I was ignorant of. Reinventing the psychodynamic wheel.
Revisiting key ideas
- Enactive Representation
- The idea that basic representation is tightly linked to motor activity and recreating past states.
- is this really the standard meaning?
- Ask Claude
- Short answer: partially, but your framing adds something non-standard.
- There are actually two distinct traditions using the term, and your outline blends them in a way that is defensible but not straightforwardly standard:
- **1. Bruner's "enactive representation" (the more common textbook meaning)**
- In Bruner's own words: "By enactive representation I mean a mode of representing past events through appropriate motor response."
- In this mode, knowledge is stored in the form of motor responses; thinking is based on physical actions, with infants learning by doing rather than by internal representation.
- This matches your outline's "tightly linked to motor activity" part well — but the "recreating past states" framing is your gloss, not exactly Bruner's language. Bruner's point is more that the motor response is the representation, not that the organism is simulating or reconstructing something.
- **2. The enactive cognition tradition (Varela, Thompson, Rosch)**
- The enactive approach to cognitive science proposed a new set of theoretical assumptions aimed at breaking the tight conceptual connection between cognition and representation. Instead, cognitive processes are deeply entangled in action. Key aspects are: (1) perception consists in perceptually guided action and (2) cognitive structures emerge from the recurrent sensorimotor patterns that allow action to be perceptually guided.
- This tradition is actually skeptical of representation as a concept —
- enactive approaches to cognitive science aim to explain cognitive processes without making any appeal to internal, content-carrying representational states.
- The tension in your outline
- Your framing — "basic representation is tightly linked to motor activity and recreating past states" — is closer to Bruner than to Varela/Thompson, but it introduces the idea of recreating past states (which sounds like motor simulation or re-enactment) that neither tradition emphasizes as the central mechanism. That idea has more kinship with theories like motor simulation (Jeannerod), embodied simulation (Gallese), or the role of mirror neuron systems.
- The activity of mirror neurons when observing another person's actions, which activates one's own motor programs, is indeed interpreted as an enactive representational process
- by some, but this is an extension, not the core definition.
- So: the "linked to motor activity" part is standard for Bruner; the "recreating past states" part is a meaningful addition your outline is making — worth flagging explicitly as a theoretical move rather than a definition.
- Schemas (in the Drescher/Piagaet sense)
- Learning schemas
- Using schemas – the imaginal
- Emotion
- as valenced reaction to imagined futures
- Conflict
- Displacement activities, ritual, non-compromise
- Chaining
- The above is a sort of primitive mechanism for anticipating future states; this theorizes (very briefly) about how those mechamisms can be stiched together for more complicated intelligent action.
- Summary
- Emotion and cognition are deeply intertwined
- Representation is grounded in experience; facts and objectivity are later developments at best
- Extensions and speculations
- Object relations theory (psychoanalysis)
- The other, the self, and Kegan stage theory
- Chaining depends on self-representation
- Complex emotions from complex self/other representations
Furthur
- The predictive moment: reverie, connection and predictive processing (relates PP to Bion's psychoanalytic theory)
- Blaise Agüera y Arcas has talked of similar things in an AI context eg in Do Large Language Models Understand Us? | Daedalus | MIT Press
The idea of having a social relationship with oneself might seem counterintuitive, or just superfluous. Why would we need to construct models of ourselves if we already are ourselves? One reason is that we are no more aware of most of what actually happens in our own brains than we are of anyone else's.