Integrating 4E cognition with science and technology studies: a framework for understanding AI applications

Abstract

The paper brings together two different theoretical strands of research, one from cognitive science, the other from Science and Technology Studies. The purpose in so doing is to uncover how cognition interrelates with socio-material practices and AI technology. An integrative framework is presented as a possible way for connecting the two strands while theorizing on their interrelations with AI.

1 Introduction

Artificial Intelligence is becoming increasing integral to human socio-material practices (Wellner, 2023). Whether such practices are instantiated as organizations, establishments or one thing else, it’s evident that the spreading of AI is just not restricted to remoted socio-practical domains (Peschl, 2024) and, furthermore, that the ‘full potential’ of the technology but must be unleashed. In reality, the possibly unrestricted applicability of Large Language Models akin to ChatGPT makes it crucial to acknowledge that AI applied sciences will find yourself being, not simply utilized by people as mere instruments, however, much more so, come to interaction with the assorted methods through which our cognition and the practices we partake in evolve over time. The objective of the current paper is to theorize on the interrelation of cognition, practices and AI technology, and to take action by participating with two radical strands of analysis, one coming from cognitive science, the opposite from Science and Technology Studies (STS). The total objective¹ is to deliver these nearer collectively and stress compatibility by presenting a framework for integrating a few of their insights.

2 Theoretical background

2.1 4E cognition

Radical approaches to cognition in contemporary cognitive science are synonymous with the so-called 4E Cognition theories. Those in favor of a radical view generally assume that cognition is extended into the agent’s surroundings, embodied in that it involves not just the brain but also the body of the agent, enacted by the agent and, finally, embedded in actual contexts; rather than being simply reducible to an intracranial phenomenon (see, e.g., Marr, 1982). This contrasts with how these in favor of so-called Representational Theory of Mind (e.g., Fodor, 1981) take into account cognition. They assume that “postulating representational (or ‘intentional’ or ‘semantic’) states is essential to the theory of cognition” (Fodor and Pylyshyn, 1988).

In pushing the 4E agenda, radical approaches differ from orthodox, representationalist approaches to cognition as they criticize the idea that cognition takes place strictly brain side and that it is bound to be based on representational content. As Cussins puts it, “A representational content is a presentation (or re-presentation) of the world in experience or in thought” (Cussins, 1992, p. 655). In this connection, there are completely different levels of radicality amongst 4E approaches. Some proponents of anti-representationalism grant that so-called illustration hungry drawback domains do exist. The existence of those domains pertains to the concept these sorts of cognition that contain overtly reflective or deliberate attitudes akin to planning or formulating propositions (cf. Clark and Toribio, 1994; Hutto and Myin, 2013) are sure to contain representations. Others, nevertheless, take the unconventional agenda a step additional by searching for to get rid of illustration starvation altogether (Degenaar and Myin, 2014; Kiverstein and Rietveld, 2018). Nevertheless, even if there’s room for disagreements in relation to extra ‘sophisticated’ sorts of cognition, what notions one ought to use for describing non-representational cognition (see, e.g., Hutto and Myin, 2013) and so on. there’s nonetheless a basic, fundamental settlement amongst proponents of 4E that cognition may be theorized with out presupposing the workings of psychological content material and representations (Gahrn-Andersen, 2021). However, what’s at present lacking from 4E approaches to cognition is that the interrelation between cognition and the unfolding of practices is mostly not pursued in a systematic trend. And that is even if it’s acknowledged that the reliance on affordance relations by way of engagements with instruments, issues, computer systems and so on. is just not solely attribute of a state of affairs, however trans-situational within the sense of being tied up with practices that are irreducible to the particularities of a state of affairs. As van Dijk and Kiverstein put it, affordances “are situated, but concretely available aspects of the sociomaterial environment to coordinate to. We can see that they function in these manifold ways if we zoom out on our practices in space and time to notice how chairs, doors, benches, paths and ponds are entangled within and across concrete situations” (van Dijk and Rietveld, 2017, p. 6). Despite not having engaged with the difficulty of trans-situationality of cognition from a constitutive perspective (one may maybe declare that Hutchins does so however descriptively), 4E theories present good insights into mechanisms that enable cognition to unfold as a located, sensible phenomenon.

2.1.1 Know-how and affordances

Know-how is important in the sense that it enables flexible adaptive behavior; both in socially saturated situations and beyond. Such knowhow lends itself to embodied expectations (which are often tacit and not previously articulated) and possibilities for actions, and it characterizes even our plain, everyday enacted perceiving (Hutto, 2005, p. 391). Indeed, such intentional directedness is considered as counting on know-how or being expert in that it entails “coordinating with multiple affordances simultaneously in a concrete situation” (van Dijk and Rietveld, 2017, p. 8). Affordances, on this view, are alternatives for actions that exist as a relation between an agent and its setting (Gibson, 1979). In different phrases, cognition depends each on agent-specific expertise and capacities in addition to options within the materials world; the world thus come to play a constitutive function to cognition.

2.1.2 Practical understanding

In building on key insights from the early Heidegger, Noë (2015) makes the declare that our notion and sensible dealings extra usually are guided by our sensible understanding. Noë argues: “What makes [understanding] practical […] is that it is the gearing in or putting to work of one’s understanding in the absence of any call for, or even space for, reflection or judgement” (2015, p. 3). In different phrases, it occurs within the ‘absence of deliberation’ (2015, p. 4) however however permits us to discriminate virtually related affordances from irrelevant ones. Moreover, it permits for one thing to afford X in a given state of affairs and not Y or Z. Indeed, sensible understanding may be seen because the enabler of know-how in that it permits know-how for use ‘correctly’ and, thus, for brokers to tacitly undertake their expertise by way of engagements with affordances, not simply to what the state of affairs affords, but in addition, to what makes affordances virtually entangled with different conditions relative to the apply in query (see, Hutchins, 1995 for quite a few examples of such understandings at play in a tightly coupled group or Heidegger, 2010 for a extra technical argument on how issues stand in sensible relevance to different issues).

2.2 Performativist STS

Just as in the case of cognitive science, contemporary Science and Technology Studies (STS) is characterized by an internal, paradigmatic shift. Indeed, some proponents of radical cognition have been stressing a degree of compatibility (De Jesus, 2018). Also, on this case, the break up consists between, on the one hand, those that argue in favor representational approaches to the research of scientific practices and technology-use extra usually and, on the opposite, these advocating in favor of a performative strategy. The representational idiom is characterised by a indifferent view of technology use, the place the human agent is seen as separate from the world. This detachment entails that the agent is ready to motive or cognize about worldly affairs in an goal method; mainly observing states of affairs from a distance (Jensen, 2004) Performativists akin to Haraway (1991), Pickering (1995), Latour (1999) and many others, however, emphasize the significance of sensible ontologies. Such ontologies are thought of as options to certainties of information (i.e., epistemology) and psychological representations. Moreover, proponents of performativism argue that we should always not impose an analytical distinction between social info and cognitive info because the two are enmeshed within the outset (cf. Jensen, 2004, p. 237). In phrases of thematizing the trans-situationality of practices, we discover not less than two helpful notions; not solely for exploring sensible exercise as one thing that unfolds over time, but in addition helpful notions bringing 4E insights collectively with performativist ones.

2.2.1 Activity trails

Based on the works of Cussins (1992), exercise trails are minimally outlined as patterns or ‘forms of guidance’ that transfer by way of environments of exercise. They not solely “lend stability” to classes akin to ‘the political’, ‘the technological’ and ‘the social’ (Jensen, 2017, p. 631) but in addition assist to flee the traps of humanist approaches. Simply, they don’t take the human agent as their analytical floor. Consequently, they permit “decentred analyses in which subject and object formations are emergent outcomes of material—relational processes” (Jensen, 2017, p. 631). They are additionally used to counter claims that performativist positions entail “radical relativism” by embracing sensible ontologies as a substitute of objectifying epistemologies (Gad et al., 2015). Indeed, Jensen appeals to exercise trails in infrastructure configurations, exhibiting that “they help recreate not only the ‘objective’ space of the city, but also the ‘subject positions’ of those who depend on these infrastructural arrangements” (Jensen and Morita, 2017, p. 8). Activity trails join what’s goal and subjective, and they transcend focusing solely on the pondering individual as the middle of actions and occasions.

2.2.2 Conjunctions

‘Conjunctural events’ or just ‘conjunctions’ describe how novelties come about “out of various contingent combinations of heterogeneous instruments, temporal and spatial dispositifs, procedures and techniques; something else than what caused it, which has its own specific historical trajectory; […] something which does not bring together already existing objects, subjects and social groupings” (Hennion and Grenier, 2000, p. 9). But the relevance of conjunctions goes past inventive practices to mainly contain any prevalence the place completely different socio-material entanglements generate novelties together with, for occasion, as options to deltas in disaster such because the Mekong and Chao Phraya (Jensen and Morita, 2020).

3 Theoretical integration

AI technologies are world-involving (Peschl, 2024) within the sense that we make their outcomes affect or be influenced by the unfolding of on a regular basis practices. AI can assist or exchange actions, serving as affordances to a apply. For occasion, one can ask ChatGPT to generate a recipe, draft a program for a convention or write a speech. Based on the output generated by the AI, the human consumer can resolve to enact features hereof, not simply onto their environment, but in addition in manners whereby they tie in with specific options of an present apply (Sadek et al., 2024). The technology allows customers to dump cognitive effort (Chemero, 2009) by rushing up duties, bettering efficiency and planning, making some duties pointless, and even creating new ones. In quick, they permit for completely different relationalities between brokers and features of their environments (Figure 1).

We have interaction with AI applied sciences in ways in which mirror our expert intentionality. Consider the latest development of individuals utilizing generative AI to create motion figures of themselves or others. This phenomenon could have originated by accident—maybe somebody was casually experimenting and discovered the outcomes intriguing sufficient to share on social media. Countless different inventive creations fail to realize related viral impression. Regardless, this development can be utilized to showcase the intentionality that underlies inventive engagements with LLM-based applied sciences. Such intentionality is rooted in a twofold construction: know-how and sensible understanding. In the case of the motion figures, it includes not solely familiarity with immediate engineering—the talent of crafting efficient prompts—but in addition a sensible grasp of what qualifies as an ‘action figure’. Furthermore, it’s the similar form of sensible understanding that permits us to critically consider AI outputs (for a concrete instance, see Gahrn-Andersen, 2025).

When a part of a broader development, the customers make use of their experience not solely in working the AI system but in addition in integrating its outputs meaningfully into actual or digital contexts. Generative fashions like ChatGPT or DALL·E 3 can be utilized for inventive expression or just to go the time. This highlights the truth that their outputs don’t essentially have rapid sensible utility. Instead, they could be inventive, exploratory, or idiosyncratic in nature. AI applied sciences can thus be used conjuncturally—creatively or in methods that aren’t immediately tied to present sensible configurations. While these makes use of could not feed into established practices, besides maybe these associated to artwork (cf. Hennion and Grenier, 2000; Chatterjee, 2022), they nonetheless maintain potential significance. Conjunctural makes use of can function websites of innovation, bringing novelty to established practices.

Although the creation of an motion determine utilizing AI needn’t afford rapid sensible use, the latest development reveals that it could actually serve completely different functions—akin to self-promotion on platforms like Instagram or Facebook, or in additional skilled contexts: on LinkedIn. The reputation of making self-crafted motion figures as a ‘social media trend’² reveals how this phenomenon interrelates with posting practices on completely different social media platforms. This displays the trans-situationality of the apply: not solely is it re-enacted throughout completely different platforms, nevertheless it additionally persists and evolves throughout these completely different contexts and over time. While the outcomes are partially decided by how the AI features internally, their perceived high quality and relevance additionally rely on the consumer’s platform-specific data and disposition—how and the place the generated content material is shared. Finally, it’s essential to notice that AI use evolves so quickly that ostensive descriptions—specific, shareable directions—for find out how to immediate LLMs to generate particular outputs, akin to motion figures, are actually available.³ The know-how required to generate an motion determine of a specific individual has shifted from being predominantly idiosyncratic and relative to the actions of specific people to turning into basic. Prompt engineering has matured into a recognizable apply, full with pointers and knowledgeable suggestions. As a consequence, the ‘practice-constitutive enactment’ of producing such content material has been explicated and made ostensible, making it replicable and teachable. However, if one doesn’t need to hassle with nice tuning prompts for ChatGPT or DALL·E 3, one can as a substitute use Flux AI’s Action Figure Generator which is designed particularly for the aim of producing motion figures. Here, one merely gives fundamental details about the sort of determine one intends to create a illustration of, with out having to inform the AI what constitutes an motion determine, from what angle and so on. Thus, one doesn’t must have refined prompting expertise as a way to make an motion determine.

Considering how Flux AI promotes its providers, it turns into clear that AI-generated materials, akin to photographs of custom-made motion figures, can assist a wide selection of sensible applications and probably catalyze various types of engagement, constituting completely different exercise trails. For instance, their web site highlights completely different sensible domains akin to content material creation by influencers, present giving, amassing, and use by small companies to design mockups for {custom} merchandise or promotional collectibles.⁴ In basic phrases, the agent’s capacities and tendencies constitutes what an AI system could provide in a given context, lending assist to decision-making (Eisbach et al., 2023) and triggering new paths of motion. These new exercise trails not solely form how the apply evolves over time but in addition affect the consumer’s personal expertise and understanding, permitting these to develop as properly (Benvenuti et al., 2023).

4 Conclusion

The paper shows that theoretical notions from 4E Cognition theories and performativist STS can be brought together in a fruitful interplay for the purpose of exploring how cognition and practices interrelate in the use of AI. The proposed framework can inspire future research in both camps. Specifically, it can inspire proponents of 4E cognition theories to extend their concepts into the realm of AI and human socio-material practices more general. For proponents of performativist STS, on the other hand, it effectively brings the cognitive back into the loop by demonstrating that cognitive phenomena can be treated in non-representationalist (or, simply: performativist) terms and that cognitive mechanisms such as know-how and practical understandings relate to AI-use in the enactment of practices.

References

• BenvenutiM.CangelosiA.WeinbergerA.MazzoniE.BenassiM.BarbaresiM.et al. (2023). Artificial intelligence and human behavioral development: a perspective on new skills and competences acquisition for the educational context. Comput. Hum. Behav.148:107903. doi: 10.1016/j.chb.2023.107903

• ChatterjeeA. (2022). Art in an age of artificial intelligence. Front. Psychol.13:1024449. doi: 10.3389/fpsyg.2022.1024449

• ChemeroA. (2009). Radical embodied cognitive science. Cambridge, MA: MIT Press.

• ClarkA.ToribioJ. (1994). Doing without representing. Synthese101, 401–431. doi: 10.1007/BF01063896

• CussinsA. (1992). Content, embodiment, and objectivity: the theory of cognitive trails. Mind101, 651–688. doi: 10.1093/mind/101.404.651

• De JesusP. (2018). Thinking through enactive agency: sense-making, bio-semiosis and the ontologies of organismic worlds. Phenomenol. Cogn. Sci.17, 861–887. doi: 10.1007/s11097-018-9562-2

• DegenaarJ.MyinE. (2014). Representation-hunger reconsidered. Synthese191, 3639–3648. doi: 10.1007/s11229-014-0484-4

• EisbachS.LangerM.HertelG. (2023). Optimizing human-AI collaboration: effects of motivation and accuracy information in AI-supported decision-making. Comp. Hum. Behav. Artif. Hum.1:100015. doi: 10.1016/j.chbah.2023.100015

• FodorJ. A. (1981). Representations: Philosophical essays on the foundations of cognitive science. Cambridge, MA MIT Press.

• FodorJ. A.PylyshynZ. W. (1988). Connectionism and cognitive architecture: a critical analysis. Cognition28, 3–71. doi: 10.1016/0010-0277(88)90031-5

• GadC.JensenC. B.WinthereikB. R. (2015). Practical ontology: worlds in STS and anthropology. Natureculture, 3, 67–86.

• Gahrn-AndersenR. (2021). Reckoning with the stakes in overcoming representation-hungry problem domains. Acta Anal.36, 517–532. doi: 10.1007/s12136-021-00462-7

• Gahrn-AndersenR. (2025). Beyond symbol processing: the embodied limits of LLMs and the gap between AI and human cognition. AI Soc.40, 3105–3107. doi: 10.1007/s12136-021-00468-y

• GibsonJ. J. (1979). The ecological approach to visual perception. Boston: Houghton Mifflin Company.

• HarawayD. (1991). Simians, cyborgs, and women: The reinvention of nature. London: Routledge.

• HeideggerM. (2010). Being and time (StambaughJ., Trans.). Albany, NY: SUNY Press.

• HennionA.GrenierL. (2000). Sociology of artwork: New stakes in a post-critical time. Available on-line at: https://www.researchgate.net/publication/32230944_Sociology_of_Art_New_Stakes_in_a_Post-Critical_Time (Accessed November 11, 2024).

• HutchinsE. (1995). Cognition in the wild. Cambridge, MA: MIT Press.

• HuttoD. D. (2005). Knowing what? Radical versus conservative enactivism. Phenomenol. Cogn. Sci.4, 389–405. doi: 10.1007/s11097-005-9001-z

• HuttoD. D.MyinE. (2013). Radicalizing enactivism: Basic minds without content. Cambridge, MA: MIT Press.

• JensenC. B. (2004). A nonhumanist disposition: on performativity, practical ontology, and intervention. Configurations12, 229–261. doi: 10.1353/con.2006.0004

• JensenC. B. (2017). Pipe dreams: sewage infrastructure and activity trails in Phnom Penh. Ethnos82, 627–647. doi: 10.1080/00141844.2015.1107608

• JensenC. B.MoritaA. (2017). Introduction: infrastructures as ontological experiments. Ethnos82, 615–626. doi: 10.1080/00141844.2015.1107607

• JensenC. B.MoritaA. (2020). Deltas in crisis: from systems to sophisticated conjunctions. Sustain. For.12:1322. doi: 10.3390/su12041322

• KiversteinJ. D.RietveldE. (2018). Reconceiving representation-hungry cognition: an ecological-enactive proposal. Adapt. Behav.26, 147–163. doi: 10.1177/1059712318772778

• LatourB. (1999). Pandora’s hope: Essays on the reality of science studies. Cambridge, MA: Harvard University Press.

• MarrD. (1982). Vision. New York: W. H. Freeman.

• NoëA. (2015). “Concept pluralism, direct perception, and the fragility of presence” in Open MIND: 27(T). eds. MetzingerT.WindtJ. M. (Frankfurt am Main: MIND Group).

• PeschlM. F. (2024). Human innovation and the creative agency of the world in the age of generative AI. P.S. and Soc.2, 49–76. doi: 10.1177/27538699241238049

• PickeringA. (1995). The mangle of practice: Time, agency and science. Chicago: University of Chicago Press.

• SadekM.KallinaE.BohnéT.MougenotC.CalvoR. A.CaveS. (2024). Challenges of responsible AI in practice: scoping review and recommended actions. AI Soc.40, 199–215. doi: 10.1007/s00146-024-01880-9

• van DijkL.RietveldE. (2017). Foregrounding sociomaterial practice in our understanding of affordances: The skilled intentionality framework. Front Psychol, 7. doi: 10.3389/fpsyg.2016.01969

• WellnerG. (2023). Material hermeneutic of digital technologies in the age of AI. AI and Soc. 38, 2159–2166. doi: 10.1007/s00146-020-00952-w