Clark’s work explores a number of disparate but interrelated themes. Many of these themes run against established wisdom in cognitive processing and representation. According to traditional computational accounts, the function of the mind is understood as the process of creating, storing and updating internal representations of the world, on the basis of which other processes and actions may take place. Representations are updated to correspond with an environment in accordance with the function, goal-state, or desire of the system in question at any given time. Thus, for example, learning a new route through a maze-like building would be mirrored in a change in the representation of that building. Action, on this view, is the outcome of a process which determines the best way to achieve the goal-state or desire, based on current representations. Such a determinative process may be the purview of a Cartesian "central executive" or a distributed process like homuncular decomposition.
In contrast to traditional models of cognition, which often posit the one-way flow of sensory information from the periphery towards more remote areas of the brain, Clark has suggested a two-way "cascade of cortical processing" underlying perception, action, and learning. The concept of predictive processing lies at the heart of this view, wherein top-down predictions attempt to correctly guess or "explain away" bottom-up sensory information in an iterative, hierarchical manner. Discrepancies between the expected signal and actual signal, in essence the "prediction error," travel upward to help refine the accuracy of future predictions. Interactions between forward flow of error (conveyed by "error units") and backward flow of prediction are dynamic, with attention playing a key role in weighting the relative influence of either at each level of the cascade (dopamine is mentioned as "one possible mechanism for encoding precision" with regard to error units). Action (or action-oriented predictive processing) also plays an important role in Clark's account as another means by which the brain can reduce prediction error by directly influencing the environment. To this, he adds that "personal, affective, and hedonic" factors would be implicated along with the minimization of prediction error, creating a more nuanced model for the relationship between action and perception.
According to Clark, the computational model, which forms the philosophical foundation of artificial intelligence, engenders several intractable problems. One of the more salient is an information bottleneck: if, in order to determine appropriate actions, it is the job of the mind to construct detailed inner representations of the external world, then, as the world is constantly changing, the demands on the mental system will almost certainly preclude any action taking place. For Clark, we need relatively little information about the world before we may act effectively upon it. We tend to be susceptible to "grand illusion", where our impressions of a richly detailed world obscures a reality of minimal environmental information and quick action. We needn't try to reconstruct the detail of this world, as it is able to serve as its own best model from which to extract information "just in time".
Clark’s writings also focus on the concept of transhumanism, most prevalent in his work, Natural-Born Cyborgs which explores the progressing incorporation of human biology and technological implants. Through a series of contemporary technological studies and an evaluation of the cyborg figure in pop-culture, Clark maps out a perception of the cyborg as a reality. This is not necessarily to show what humanity is to become from biologically implanted technology, but rather to explore where humanity is now with said technology. In his own words, humans are "creatures whose minds are special precisely because they are tailor-made for multiple mergers and coalitions." He elaborates this as he describes his body as an "electronic virgin" untouched by technology, but gradually over time technology will become intertwined with his biology. Whether that incorporation will be as mundane as the use of eyeglasses or something more advanced such as a new auditory prosthesis, he believes the merger of technology and biology is inevitable and present.
Clark is perhaps most well known for his defence of the extended mind hypothesis. According to Clark, the dynamic loops through which mind and world interact are not merely instrumental; the cycle of activity that runs from brain through body and world and back again is what constitutes cognition. The mind, on this account, is not restricted to the biological organism but extends into that organism's environment. An example is carrying out a mathematical task. One person may complete the task solely in their head, while another completes the task with the assistance of paper and pencil. By Clark’s parity principle, there is no reason to count these means as different so long as the results are the same. The process of cognition in the second case involves paper and pencil, so the conception of mind appropriate to the person involved must include these items.
Clark concedes that, in practice, the criterion of "equal efficiency" required by the parity principle is seldom met. He nonetheless proposes that the boundary of "skin and skull" is arbitrary and cognitively meaningless. If, in the example above, the paper and pencil used by the second person becomes a virtual paper-and-pencil visible on a monitor and controlled by a silicon chip implanted in the head, the similarity between the two situations becomes clearer.
Clark foresees the development of cognitive prosthetics, or "electronic brain enhancements" ("EBEs"), as only the next logical step in the human mind’s natural integration with technology. Clark’s research interests also include wetwiring and other human-electronic integration experiments, as well as technological advances in immediate human communication and their utilization in society.
To further illustrate extended cognition, Clark uses two anecdotal examples: Inga and Otto, the latter being a patient of Alzheimer’s disease. Both desire to visit a museum and must remember the address. For Inga, this is a matter of simple memory recall and accessing her belief in where she remembers the given location of the museum. Due to his condition, Otto relies on the use of written records in a notebook in order to recall his memory, as the disease makes it difficult for him to have a finite belief on the museum's location without it. This notebook in a way becomes his livelihood, functioning as a memory bank to accommodate his condition. As Clark puts it, "his notebook plays the role usually played by a biological memory" effectively becoming a part of his extended mind. The goal of describing the thought processes of Inga and Otto is to show multiple variations of extended cognition and how it functions differently among individuals depending on their mental biology. Clark claims the key aspect that unites the cognitive process of the two individuals is that each has the initial belief on where the museum is located. Although the way the verification of belief differs in practice between the two, they are still able to come up with a solution on the museum’s location despite one using a source of external storage. In a later work, Clark addresses any potential criticism on the process of Otto’s system of belief by stating that Otto’s impulse to consult his notebook is a similar process to that of Inga consulting her recalled memory. Clark also elaborates on the importance of environment when using extended cognition. He believes that the way one formalizes "beliefs" is, "constituted partly by features of the environment" if they are playing the desired role of "driving cognitive processes.” This summarizes his belief that the mind, in fact, can extend to the surrounding environment.
Although the comparisons between Otto and Inga may seem distant when judging the power of extended cognition, Clark attempts to gap a more similar comparison between the two by claiming that if Otto were to lose his notebook it would be the same as Inga losing a thought from her consciousness. He also suggests the concept of Otto’s notebook being an extension of himself, having it take on a more material role suggesting that it may be equal to the way one treats a physical appendage to their body. The notebook in a way becomes a "fragile biological limb or organ" one that Otto may feel to protect from harm's way.
Supporters of extended cognition have used Clark’s comparison between Inga and Otto as a champion analogy to support the theory of the internal mind extending into the environment. In an article on the argument of extended cognition, Erik Myin makes note of this group, "a close causal coupling between persons and environments," such as Otto and his notebook, "can license the conclusion that the mind spreads into the environment." His also writes that some followers use Clark’s argument of, "external elements" playing a role that could be seen as, "cognitive if played by something internal to a person." A good reference to this would be how Otto's notebook is used for cognitive verification.
The Extended Mind has been met with criticism from scholars challenging extended cognition. The claim is that Clark’s theories lack a differentiation of biological memory and external storage of thought. This is mainly addressed to the cognitive comparison between Otto and Inga, and how their thought processes are viewed as equal an engagement of extended thought given Otto's condition. When he uses his notepad, it is necessary reliance, as these records will help him to remember his thoughts. However, there is an argument questioning if this follows Clark's description of extended cognition as it requires a running cycle of cognitive connectivity that returns to the body.
In a critique of Clark’s Supersizing the Mind, Lawrence Shapiro and Shannon Spaulding, of The University of Wisconsin-Madison, touch upon this criticism from scholars who challenge extended cognition. They describe the argument against Clark being that "...cognitive routines involving biological memory and those involving external stores of information are importantly different." Essentially they are claiming that the way someone stores memory externally is separate from their biological cognition. For example, if Inga were to use a notepad, it would function as an extension of her compression as her thought process would follow Clark’s description of dynamic loops that return to the mind in a cycle. However, when Otto uses the notepad has a way to retain his memory, whatever thoughts he records will not follow back and will become disconnected from his biological mind. In this respect, parts of Otto’s mind would essentially "exist outside his brain," meaning that the comparison between Inga and Otto may not be reliant as they appear to function differently in Clark's own theory.
This criticism of Otto and Inga is addressed by Clark himself in Supersizing the Mind:
“[the] claim was not that the processes in Otto and Inga are identical, or even similar, in terms of their detailed implementation. It is simply that, with respect to the role that the long-term encodings play in guiding current response, both modes of storage can be seen as supporting dispositional beliefs. It is the way the information is poised to guide reasoning … and behavior that counts. (p. 96)”
He argues that extended cognition was never meant to be a universal "fine-grained similarity" between all minds. He believes that such a "fine-grained similarity" is not necessary when analyzing extended consciousness, suggesting it differs depending on an individual's behavior and their given environment.