Transactive memory

The relation between environmental and brain-based resources in the context of memory was first studied in the context of social memory (relying on another individual agent to perform a memory task), also labelled as transactive memory (Wegner Reference Wegner, Mullen and Goethals1987). Transactive memory defines a shared system of encoding, storing, and retrieving information, where individuals delegate certain knowledge to other group members based on their expertise. This process relies as much on the memory of content as on the memory of sources and means of accessing this content, thus creating a distributed cognitive network (Rabb et al. Reference Rabb, Fernbach and Sloman2019; Nagam Reference Nagam2023). Everyone memorizes not only content, but also who knows what and how to access this information within the group.

Several experiments have shown, for example, that when two people are exposed to the same information, they tend to selectively forget what they believe their partner will remember better, based on implicit knowledge of their partner’s areas of expertise (Rabb et al. Reference Rabb, Fernbach and Sloman2019). Transactive memory processes take place within a small group, like a couple, a family, or amongst co-workers. In Western societies, women tend to take the load of birthday dates or child-related information. The gendered way to offload and distribute information should be more examined in future studies on EXT. Transactive Memory Systems are genuinely useful, and the benefits are multiple. They permit an expansion of knowledge and a reduction of cognitive load for the individuals (Wegner Reference Wegner, Mullen and Goethals1987). Collective memory is also ‘crucial to humans as a social species’ (see Michaelian and Sutton Reference Michaelian and Sutton2013; see also Gensburger and Clavert Reference Gensburger and Clavert2024; Rabb et al. Reference Rabb, Fernbach and Sloman2019). They constitute structured systems, with several interdependent dimensions. They are based first on coordination and communication processes, which make it possible to know where information is stored in the group (Wegner Reference Wegner, Mullen and Goethals1987).

Transactive memory also involves metacognitive knowledge: knowing who knows what and how to access that knowledge. Another fundamental aspect is the shared spatiotemporal environment: information is often only accessible within a social interaction, during a collaborative retrieval task. As Sutton et al. (Reference Sutton, Harris, Keil and Barnier2010) point out, many memories are not mobilized autonomously but require concrete interaction with others or the environment. If a group member is absent, the information may become inaccessible. Interdependence between group members and a dynamic flow of information between partners are central characteristics. Transactive memory is a property of the group and is not limited to a simple transfer of information. Studies show that interaction between group members has causal power. Among older couples in particular, it is the exchange itself that gives rise to information, not its prior possession by either partner (Sutton et al. Reference Sutton, Harris, Keil and Barnier2010). Thus, transactive memory constitutes a strong argument for thinking of memory as an essentially distributed phenomenon.

Technological environment and cognitive offloading strategy

Although it is not the main focus of this paper, this detour through social memory is necessary. It has been argued, in fact, that digital technologies can be used as a kind of transactive memory partner (Sparrow et al. Reference Sparrow, Liu and Wegner2011; Storm et al. Reference Storm, Stone and Benjamin2016). External memory technologies consist of ‘digital systems and services we use to record, store, and access digital memory traces to augment, re-use, or replace organismic systems of memory’ (Clowes Reference Clowes2013). We can offload information on physical support, like a sheet of paper or a note in your smartphone. It is common to take a screenshot of a fun fact, to take a photo of a poster related to an upcoming event, and so on.

It appears that human subjects are capable of forming transactive memory systems not only with other individuals, but also with technological artefacts such as the Internet, which then become an environmental way for storing and retrieving information. This phenomenon echoes the thesis defended by Clark and Chalmers (Reference Clark and Chalmers1998), according to which the technological environment can constitute extensions of the mind. However, the effects and mechanisms by which such distributed memory is organized in the interaction with digital technologies and its relationship with semantic memory are little studied and remain poorly understood or even unknown (Nagam Reference Nagam2023). It seems necessary to better understand the way in which digital technology comes to function as an extension of our cognitive system, or even as a partner with which users develop expectations, routines for accessing information, or even a cognitive dependence. This raises crucial questions for the philosophy of memory (Sutton et al. Reference Sutton, Harris, Keil and Barnier2010; Heersmink Reference Heersmink2016). Technological memory is seen as an aid in order to supply the deficiency of brain-based memory or facilitate memory accessibility (Silva et al. Reference Silva, Pinho, Macedo and Moulin2016). It is a genuine cognitive process, proceeding in interdependence and interrelations between technological memory and neural processes (Bell and Gemmel Reference Bell and Gemmel2009; Sutton et al. Reference Sutton, Harris, Keil and Barnier2010).

The Google effect

The Google effect, named by Sparrow et al. (Reference Sparrow, Liu and Wegner2011), has been widely cited as evidence that digital tools alter human memory by encouraging cognitive offloading. Many comparisons have been made with the concept of transactive memory and the Google effect (Ward Reference Ward2013; Wegner & Ward Reference Wegner and Ward2013; Clowes Reference Clowes, Bernecker and Michaelian2017; Heersmink and Sutton Reference Heersmink and Sutton2020; Eliseev and Marsh Reference Eliseev and Marsh2021). Sparrow et al. (Reference Sparrow, Liu and Wegner2011) claim that social transactive memory has a new manifestation supported by constant online access and search engines. The original study tested participants’ memory for trivia statements saved in computer folders. The core finding, that participants remembered where information was stored better than the information itself (Sparrow et al. Reference Sparrow, Liu and Wegner2011), has been interpreted through the lens of EXT, suggesting a growing symbiosis between human agents and digital tools. However, the study’s simplicity, the absence of true interactivity, and limited ecological validity cast doubt on its conceptual scope. Later attempts at replication have yielded mixed results: while Camerer et al. (Reference Camerer, Dreber, Holzmeister, Ho, Huber, Johannesson, Kirchler, Nave, Nosek, Pfeiffer, Altmejd, Buttrick, Chan, Chen, Forsell, Gampa, Heikensten, Hummer, Imai, Isaksson, Manfredi, Rose, Wagenmakers and Wu2018) failed to replicate the original findings, Schooler & Storm (Reference Schooler and Storm2021) succeeded under specific conditions of perceived reliability, underlining the importance of trust in technological resources as a condition for cognitive offloading. Additional phenomena such as the saving-enhanced memory effect (Eliseev and Marsh Reference Eliseev and Marsh2021; Gilbert et al. Reference Gilbert, Boldt, Sachdeva, Scarampi and Tsai2023) further support the functional and adaptive nature of such offloading. Empirical evidence supports the idea that cognitive offloading can enhance performance by redistributing cognitive effort, as participants tend to remember more new information (like a word list) when earlier material (like another word list) is considered as safely stored externally (Eliseev and Marsh Reference Eliseev and Marsh2021).

The Google effect also refers to the routine use of search engines as part of our memory strategies. ‘Immediate access to information provided by search engines affects memory encoding and retrieval processes (…) people have less memory for specific information but more for where to find it’ (Musa and Bakkara Reference Musa and Bakkara2023). This conception is based on another experiment led by Sparrow et al. (Reference Sparrow, Liu and Wegner2011) known as the Google Stroop experiment (Schacter Reference Schacter2022) (Sparrow et al. (Reference Sparrow, Liu and Wegner2011) used a modified Stroop task, suggesting that Internet-related terms prime when individuals face difficult knowledge questions.). As illustrated by Carter & Gordon (Reference Carter and Gordon2017), one may retrieve online information during a conversation and mistakenly feel knowledgeable, despite relying on external sources. This blurs the line between personal knowledge and accessible online data (Ward Reference Ward2021; Schacter Reference Schacter2022). Storm et al. (Reference Storm, Stone and Benjamin2016) found that participants who initially used Google to answer questions were more likely to continue using it, even when it became harder to access. Google use thus becomes a habitual, almost invisible strategy, due to its speed, fluency, and transparency (Menary Reference Menary2012; Heersmink and Sutton Reference Heersmink and Sutton2020). The constant availability of a vast online database not only reshapes memory practices but also inflates users’ epistemic confidence, leading them to overestimate their own knowledge (Rabb et al. Reference Rabb, Fernbach and Sloman2019; Ward Reference Ward2021; Nagam Reference Nagam2023). Ward (Reference Ward2021) notes that search engines often anticipate our queries, creating a sense of seamless cognitive extension. Users may then conflate information retrieved online with internal memory, revealing the increasing permeability between brain and technology.

All of this would be worth extending to interactions with generative artificial intelligence (AI). There are currently only a few empirical studies on the impact of generative AI on semantic memory. Some suggest it could be integrated into EXT and replace current offloading strategies (Grinschgl and Neubauer Reference Grinschgl and Neubauer2022; Clark Reference Clark2025). Several recent studies, however, have begun to explore the impact of LLMs on learning, and thus indirectly on memory (Darvishi et al. Reference Darvishi, Khosravi, Sadiq, Gašević and Siemens2024; Deng et al. Reference Deng, Jiang, Yu, Lu and Liu2025; Ennion and McLellan Reference Ennion and McLellan2025; Kreijkes et al. Reference Kreijkes, Kewenig, Kuvalja, Lee, Vitello, Hofman, Sellen, Rintel, Goldstein, Rothschild, Tankelevitch and Oates2025). The advent of LLMs raises both concerns and enthusiasm (Deng et al. Reference Deng, Jiang, Yu, Lu and Liu2025). Findings remain highly variable, sometimes even contradictory (Deng et al. Reference Deng, Jiang, Yu, Lu and Liu2025; Ennion and McLellan Reference Ennion and McLellan2025). Yet, meta-analyses suggest that, in educational contexts (that is on student population), LLMs may improve student autonomy; increase, or at least support, students’ metacognitive resources and strategies (Darvishi et al. Reference Darvishi, Khosravi, Sadiq, Gašević and Siemens2024); improve academic performance and learning, positively impact students’ affective states and motivation, reduce cognitive load (Deng et al. Reference Deng, Jiang, Yu, Lu and Liu2025); improve productivity and cognitive engagement (Ennion and McLellan Reference Ennion and McLellan2025). Conversely, concerns remain that LLMs may undermine deep cognitive processing, encourage superficial practices (Kreijkes et al. Reference Kreijkes, Kewenig, Kuvalja, Lee, Vitello, Hofman, Sellen, Rintel, Goldstein, Rothschild, Tankelevitch and Oates2025), and weaken critical thinking (Deng et al. Reference Deng, Jiang, Yu, Lu and Liu2025; Kreijkes et al. Reference Kreijkes, Kewenig, Kuvalja, Lee, Vitello, Hofman, Sellen, Rintel, Goldstein, Rothschild, Tankelevitch and Oates2025). Overall, however, the available evidence is cautiously optimistic, while acknowledging significant limitations such as the small number of studies, lack of diversity, the absence of developmental populations, and possible novelty effects. An important open question concerns the long-term consequences of externalizing cognitive resources during the developmental phases of cognitive abilities (Hernández-Orallo Reference Hernández-Orallo2025).

Photo-taking impairment

Experimental evidence shows the active process of taking photos also seems to involve an automatic offloading with respect to the transparency-in-use conditions required by EXT (Transparency is a commonly accepted criterion in the literature for considering cognition as extended, and is sometimes associated with notions of fluidity or fluency [Smart et al. Reference Smart, Andrada and Clowes2022]. It can be understood along two dimensions. First, informational transparency refers to the readability and intelligibility of information: it should be easily accessible and readily interpretable by the user. Second, procedural transparency concerns the ease with which environmental resources are used, such that the agent does not need to exert special effort to engage with them; instead, they are employed in a manner comparable to organic cognitive resources. A classic example is the blind person’s cane [Merleau-Ponty Reference Merleau-Ponty1945], which, like other cognitive artefacts, can be incorporated into the body schema and function as part of the system through which individuals act in their environment [Heersmink Reference Heersmink2017]. Speed is also frequently identified as a relevant criterion in the EXT literature. External resources must allow sufficiently rapid access to information in order to support cognitive performance in real time. This requirement highlights potential limitations for some artefacts: for example, a photo album could qualify as external memory, but its utility is reduced by slow retrieval [as finding a specific image can be time-consuming]). This automatic offloading is also known as the photo-taking-impairment effect. Basically, this means that taking photos leads to the forgetting of details. ‘Photographed objects were less well-remembered than observed objects’ (Henkel Reference Henkel2014; Soares and Storm Reference Soares and Storm2018). We should say that photographed objects are less well-remembered internally. If participants are allowed to look at the picture they took, then all details of the object will be described in the most accurate way possible. The vocabulary used and the manner of expression in such studies reveal the dominant position of the internalist paradigm in psychology. Following Soares & Storm (Reference Soares and Storm2018), we can put forward several hypotheses to explain why people remember fewer details with their brains when they take pictures. Forgetting details can result from an active and explicit offloading strategy, which appears reliable, accurate, easy, quick, and saves cognitive load.

This phenomenon may also result from an implicit or automatic cognitive offloading strategy. Individuals tend to behave as if they have established a tacit transactive memory with their devices, processing visual information on the assumption that it will be reliably retained by the technological device and thus accessible later. This is true even if the experimenter warns that the information will not be available (Risko and Gilbert Reference Risko and Gilbert2016; Soares and Storm Reference Soares and Storm2018). This may also result from a metacognitive illusion. The simple act of photographing an object can generate a sense of familiarity or fluency in encoding, giving individuals the illusion of having already encoded the information about the object in their biological memory when, in reality, it has been stored, or believed to be stored, in the device (Soares and Storm Reference Soares and Storm2018).

Finally, we can also suppose that remembering photographed objects internally less well could result from an attentional disengagement state. If people know they can take pictures, they pay less attention to the details. However, the photo-taking-impairment effect disappears when participants are made to visually focus on a specific detail of the object by zooming in through the camera (Henkel Reference Henkel2014) or when participants are free to take pictures voluntarily (Barasch et al. Reference Barasch, Diehl, Silberman and Zauberman2017). Taken together, recent work suggests that taking a photo does not impair biological memory unequivocally. These effects seem to depend on the intentional posture and goals of the subject at the time of capture (Schacter Reference Schacter2022). This invites us to rethink memory as a dynamic process taking place in an environment that dictates individuals’ cognitive and metacognitive strategies. Much remains to be done to understand how cognitive strategies integrate technological resources and make them cooperate with the brain’s resources. It is becoming increasingly clear that the internalist paradigm is detrimental to a better understanding of human cognition and memory. Experimental research should evolve its methodology to integrate cognitive strategies used by people on daily routines, thus allowing better ecological validity.

Metacognitive strategies

The use of memory devices involves metacognitive strategies (Finley et al. Reference Finley, Naaz and Goh2018). The allocation of cognitive processes between biological brain and environmental technology adjusts according to circumstances, following an implicit assessment of the costs and benefits associated with each option (Grinschgl and Neubauer Reference Grinschgl and Neubauer2022). The literature identifies two fundamental dimensions of the relationship between agents and cognitive technology: selection and endorsement (Arango-Muñoz Reference Arango-Muñoz2013; Clowes Reference Clowes, Bernecker and Michaelian2017). The selection problem concerns the decision-making process by which an individual chooses, in a given context, to search for information in their own biological memory or in the environment. This decision is based on criteria such as speed of access, perceived reliability, familiarity with the device used, and integration into daily routines. For example, spending too much time searching through brain memories may be inefficient if the information is also available on a faster and more reliable technological medium (Arango-Muñoz Reference Arango-Muñoz2013). The endorsement problem refers to the evaluation of information before its use. Task success depends on endorsement, and a failure of endorsement and judgement can have harmful consequences from task or reasoning failure (Arango-Muñoz Reference Arango-Muñoz2013) to the death of the user (Gillett and Heersmink Reference Gillett and Heersmink2019) (As noted by Gillett and Heersmink, there are documented cases in which excessive or poorly calibrated reliance on GPS devices has led users to ignore salient environmental cues, sometimes with fatal consequences. While some incidents are merely anecdotal or humorous, others have involved users driving into the sea or into deserts, resulting in serious injury or death (Gillett & Heersmink, Reference Gillett and Heersmink2019). These cases illustrate how maladaptive integration of cognitive technologies may compromise situational awareness and increase vulnerability.).

A recurring concern in debates on extended memory is that information derived from environmental sources, once integrated into cognitive processes, may be automatically endorsed by the subject: accepted as reliable without prior critical evaluation (Arango-Muñoz Reference Arango-Muñoz2013; Schacter Reference Schacter2022). This raises questions about the subject’s vigilance concerning the content of their own beliefs, particularly when those beliefs originate from technological supports. The concern becomes more pressing in cases of deep cognitive integration with technological routines like the near-automatic use of a smartphone for information retrieval. In such cases, cognitive behaviour may no longer be guided by deliberate reasoning but rather by the automatic activation of a technological habit.

Metacognitive illusion and overconfidence

Some studies suggest that individuals come to perceive the Internet as a true extension of their cognitive faculties (Ward Reference Ward2013). The speed of accessing information via the Internet frequently exceeds that of exploring brain memory, which can lead to confusion between biological and environmental content. ‘Internet users may feel like they know everything that the Internet Knows’ (Ward Reference Ward2013). This tendency is robust, and a study of nine experiments led by Fisher et al. (Reference Fisher, Goddu and Keil2015) shows that relying on the Internet leads to conflating information available online with information available in the brain. The different studies mentioned here agree that participants use and treat Google search, Wikipedia, and more broadly the Internet, as a transactive memory partner (Ward Reference Ward2021; Nagam Reference Nagam2023). Some predict an increase in the externalization of semantic memory strategies (Clowes Reference Clowes2013; Carter and Gordon Reference Carter and Gordon2017). It appears easier to trust the Internet because its informational properties are different from those of brain memory. Technological means are faster, more accurate, more detailed, and less prone to errors than the latter (Heersmink Reference Heersmink2016; Ward Reference Ward2021). As a new externalized form of semantic memory, the Internet requires new metacognitive strategies that some argue are deficient: individuals overestimate what they know when they have a smartphone ready to use (Risko and Gilbert Reference Risko and Gilbert2016).

Access to the Internet can alter the feeling-of-knowing in two different ways. Overconfidence leads the participants to consider that they know what they found online or what they think is possible to find online (Fisher et al. Reference Fisher, Goddu and Keil2015; Risko and Gilbert Reference Risko and Gilbert2016), or conversely decreases the willingness to answer questions (Fergusson et al. Reference Fergusson, McLean and Risko2015; Risko and Gilbert Reference Risko and Gilbert2016). Whenever possible, participants tend to offload information, even when this cognitive offloading is unreliable. This could be a ‘metacognitive erroneous belief’ (Risko and Gilbert Reference Risko and Gilbert2016). Because the Google effect can give the illusion of knowledge, participants could misinterpret information. The Google search function is used daily by billions of people. For most individuals, searching with Google feels easier, faster, and more reliable than attempting to retrieve information from the brain, particularly in moments of uncertainty or during tip-of-the-tongue experiences.

A series of experiments conducted by Ward (Reference Ward2021) shows that using search engines leads to a blurring of the boundaries between external and internal knowledge, an attribution of online search results to one’s own brain memory, forgetting to have used a technological device, the assertion that the answers come from biological memory, increased confidence in one’s abilities, one’s intelligence and also in one’s own cerebral memory, or even erroneously optimistic predictions of the ability to answer questions without a technological device (Ward Reference Ward2021). One hypothesis to explain this phenomenon is that the speed of access offered by digital technology can be confused, on a subjective level, with the fluidity of internal recall. This confusion fuels a form of metacognitive illusion, in which the effectiveness of EXT is interpreted as a skill solely internal to the individual. Overconfidence is a challenge, and the relationship to knowledge and the metacognitive feelings that come into play with the use of search engines constitutes a scientific and ethical issue that needs to be explored in greater depth in the future.

Towards epistemic virtue

This blurring of the lines between information stored in the brain and information stored in our technological environment has important theoretical and ethical consequences regarding both what it means to know and the virtues associated with cognition. One might be tempted to reduce beliefs associated with cognitive offloading to metacognitive illusions (Risko & Gilbert, 206; Rabb et al. Reference Rabb, Fernbach and Sloman2019). On the other hand, if technological resources are integrated into cognitive processes and if the Internet is available, one could argue that individuals have good reason to rely on them and claim to know or remember something (Grinschgl and Neubauer Reference Grinschgl and Neubauer2022; Levy Reference Levy2023; Nagam Reference Nagam2023).

A significant portion of the knowledge that forms the basis of our beliefs and attitudes is not located in our brains but is distributed throughout a social, cultural, and technological environment. Some, like Rabb et al. (Reference Rabb, Fernbach and Sloman2019), consider that this difficulty in distinguishing between cerebral and environmental knowledge leads individuals ‘to overestimate how much they know or understand merely by participating in a community of knowledge’. Others, however, point out that the sharing of the burden of cognitive tasks, particularly the memorization of complex information, is a constant throughout the history of civilizations. This dynamic has favoured specialization, both at the social level and within interpersonal relationships (Nagam Reference Nagam2023). ‘Our epistemic dependence is so deep and so automatic, we may be entirely unaware we engage in it’ (Wilkinson & Levy Reference Wilkinson and Levy2024). It is therefore necessary to take into account the daily use of technology as external memory resources (Finley et al. Reference Finley, Naaz and Goh2018). Access to connected technologies in extended cognitive processes can have several benefits, such as problem solving, task execution, or accessing ‘up-to-date knowledge, and thus individuals do not need to rely on outdated, internally memorized information’ (Grinschgl and Neubauer Reference Grinschgl and Neubauer2022).

Levy (Reference Levy2024) challenges traditional accounts of intellectual autonomy that treat it as an individual epistemic virtue. He argues that such accounts place too much emphasis on individual self-sufficiency and fail to reflect the fundamentally social nature of human cognition. To address this limitation, he introduces the notion of intellectual interdependence, which better captures how cognitive achievements are often distributed across social and technological contexts.

The concept of intellectual autonomy often reinforces this mistaken view that cognitive excellence relies primarily on independence from environmental resources. Levy argues that good cognitive agents are those who actively participate in the division of intellectual labour, being based judiciously on their social and technological environment (Levy Reference Levy2023, Wilkinson & Levy Reference Wilkinson and Levy2024). Thinking ‘alone’ is an illusion, and taking into account extended cognitive processes is important for moving beyond the traditional internalist and individualist framework. Levy’s thesis is consistent with the idea that cognitive processes extend beyond the brain and is consistent with work on transactive memory, shared knowledge systems, and cognitive and metacognitive strategies supported by technological devices.

Complementarity

The biloading strategy illustrates the principle of complementarity defended by proponents of EXT: cerebral and environmental resources cooperate to perform cognitive tasks. The principle of complementarity belongs to the second wave of arguments in defence of EXT. The principle of complementarity shifts the defence of EXT away from functional similarity and towards the integration of heterogeneous resources. Cognitive systems may include environmental components that complement, rather than replicate, brain processes, provided that they are tightly and reciprocally integrated with the agent’s cognitive activity.

By combining different theoretical approaches, we can envision an extended memory where environmental memory would be associated with a preservationist view (faithful preservation of the past), while cerebral memory would fall under a more constructivist approach (Michaelian Reference Michaelian2012; Clowes Reference Clowes, Bernecker and Michaelian2017). While the brain reconstructs memories, technological devices preserve traces of the past. External resources compensate for the limits of biological memory; photos serve not only to remember, but also to enrich the precision of memory and to share them with others (Teijlingen et al. Reference Teijlingen, Oudman and Postma2022; Finley and Brewer Reference Finley and Brewer2024). Moreover, their use gives individuals a greater sense of cognitive control. In practice, these categories overlap and form a single cognitive system: technologies are also used to construct one’s narrative identity (Heersmink Reference Heersmink2022). Applications like Google Photos exemplify extended episodic memory. The app automatically sorts images, categorizes memories, and spontaneously suggests compilations (such as ‘Our Shared Memories’). This involuntary and contextual recall, triggered by the technological environment, integrates with natural processes of recollection, highlighting the extent to which our personal memories can emerge from hybrid systems, blending brain and technology.

Lifelogging

The rapid development of lifelogging technologies, defined as the continuous capture of personal data for the purpose of remembering (Silva et al. Reference Silva, Pinho, Macedo and Moulin2016), represents both a technological challenge and a challenge for our concept of memory. ‘Total capture’, ‘total recall’, ‘e-memory revolution’, ‘augment or even replace biological memories’: these are the promises of lots of practices covered under the umbrella of lifelogging technologies (Bell and Gemmel Reference Bell and Gemmel2009; Sellen and Whittaker Reference Sellen and Whittaker2010). Whether it involves photos, videos, or metadata (e.g., geolocation), the contemporary digital environment is tending to become an external, persistent, or even indexable memory (Donald Reference Donald, Malafouris and Renfrew2010). This is transforming not only our remembering practices, but also the very concepts of memory and identity, offering a new path to immortality through a digital avatar (Bell and Gemmel Reference Bell and Gemmel2009; Stokes Reference Stokes2021) (For further discussion of these issues, see the growing literature on so-called ‘ghostbots’, ‘griefbots’, or ‘deadbots’, that is, conversational agents trained on the digital traces left by deceased individuals (e.g., messages, emails, social media posts, and photographs). These systems are explicitly designed to simulate aspects of a person’s identity and communicative style after death, and are often discussed in relation to questions of memory persistence, digital legacy, and posthumous presence. See, for instance, Iglesias et al. (Reference Iglesias, Earp, Voinea, Mann, Zahiu, Jecker and Savulescu2025).

The widespread use of smartphones has facilitated a massive externalization of autobiographical traces (Soares and Storm Reference Soares and Storm2018), fuelling the idea that human memory could be replaced and augmented through total and faithful recall that overcomes the shortcomings of biological memory. As Clowes (Reference Clowes2013) points out, we are entering an era where the recording of daily life is reaching a historically unprecedented scale and fidelity. Memory is no longer located exclusively in the brain, but emerges from a hybrid system formed by the agent and its artefacts (Heersmink and Sutton Reference Heersmink and Sutton2020). Technological memory is distinguished by its stability, precision, and accessibility, while biological memory is dynamic, malleable, and closely linked to the narrative construction of the self. What happens to this identity plasticity if our memories are frozen in digital archives? The balance between internal reconstruction and external fidelity becomes a central issue for understanding future cognitive transformations and questioning the consequences of increasing recourse to preservation processes enabled by techno-environmental memory. Remembering everything could be as frustrating as remembering nothing (James Reference James1890). Intellectual and abstract abilities might perhaps be threatened by total memory (Liao and Sandberg Reference Liao and Sandberg2008).

Social media

The omnipresence of social media such as Instagram, Facebook, or TikTok in people’s lives does not need to be demonstrated. One might think that the strategies and effects of sharing personal experiences, particularly through photographs, are a well-studied phenomenon. However, several recent studies highlight that despite the omnipresence of social media, their effects on memory are still largely unknown and that the mechanisms by which they influence episodic memory remain unclear and lack ecological validity (Tamir et al. Reference Tamir, Templeton, Ward and Zaki2018; Stone and Wang Reference Stone and Wang2019; Wang Reference Wang2022; Talarico Reference Talarico2022; Wang & Hoskins Reference Wang and Hoskins2024; Vasquez et al. Reference Vasquez, Dockery, Karanian, Wang and Stone2025). The results of the studies are, moreover, not entirely homogeneous. For some, taking photos and sharing them online involves processes similar to the phenomena of transactive memory and cognitive offloading, leading to poorer brain memory for shared events (Tamir et al. Reference Tamir, Templeton, Ward and Zaki2018). Others have demonstrated that shared memories are, on the contrary, better remembered in the brain, leading to a consolidation of episodic memory but resulting in a phenomenon of retrieval-induced forgetting, that is to say that events not photographed or not shared are less well remembered and tend to disappear (Wang et al. Reference Wang, Lee and Hou2017; Stone and Wang Reference Stone and Wang2019). Episodes not shared on social media are also not stored in the brain. Finally, a recent study confirms the effect of cerebral consolidation of episodic memory but does not confirm the phenomenon of forgetting related events not photographed and not shared (Vasquez et al. Reference Vasquez, Dockery, Karanian, Wang and Stone2025).

The consolidation of episodic memory for shared photographs seems to be well explained by various factors, like the regular visualization of photographs of the same events (Stone and Wang Reference Stone and Wang2019), the importance of the photographed events (Eliseev and Marsh Reference Eliseev and Marsh2021), or the individual’s engagement in an active narrative process. Individuals photograph more of what seems important to them, and selective sharing on networks is integrated into the strategies of constructing narrative identity. Sharing with one’s network is to construct meaning (Wang et al. Reference Wang, Lee and Hou2017; Wang Reference Wang2022), to tell one’s story, which can be that of an idealized self (Stone and Wang Reference Stone and Wang2019). The publication of personal experiences allows one to reflect on their meaning and to strengthen one’s narrative coherence.

There is also a risk of false memories when false information is published and integrated into the self-narrative (Stone and Wang Reference Stone and Wang2019; Schacter Reference Schacter2022). Episodic memory is part of a complex dynamic between brain resources and environmental resources. Memories are not representations fixed in neurons of past events, but constructions made, consolidated, and forgotten through the repeated interactions that individuals have with their social and technological environment, which sometimes merge, as in the case of digital social networks. The displacement of memory towards social platforms is part of EXT, where the processing, storage, and meaning of memories are co-constructed with the environment.

Wearable camera

The most significant and rich empirical literature concerning externalization of autobiographical memory focuses on the SenseCam (for Sensing and Camera) developed by Microsoft Research (Most studies use SenseCam, but few studies also use Google Clip, narrative clip, or FrontRaw camera). The SenseCam is a wearable camera worn around the neck that takes photos without any intervention of the user (Hodges et al. Reference Hodges, Williams, Berry, Izadi, Srinivasan, Butler, Smyth, Kapur and Wood2006). This passivity process is crucial for memory, especially for people with cognitive impairments, because an automatic and passive process does not require cognitive ability to be used. This wearable camera was designed with a wide-angle lens and takes photos automatically, either according to the time elapsed since the last image or triggered by sensors. From the outset, SenseCam was thought of as a memory aid destined for patients with memory disorders, and the process of reviewing digital records of day-to-day events was supposed to stimulate memory (Hodges et al. Reference Hodges, Williams, Berry, Izadi, Srinivasan, Butler, Smyth, Kapur and Wood2006). This device is a cognitive prosthesis (Silva et al. Reference Silva, Pinho, Macedo and Moulin2013). The constant recording enabled by SenseCam-like devices is radically different from brain-based memory, which has selective encoding (Doherty et al. Reference Doherty, Pauly-Takacs, Caprani, Gurrin, Moulin, O’Connor and Smeaton2012).

Reviews of the literature show that SenseCam functions as a memory aid (Doherty et al. Reference Doherty, Moulin and Smeaton2010; Silva et al. Reference Silva, Pinho, Macedo and Moulin2013; Silva et al. Reference Silva, Pinho, Macedo and Moulin2016). It seems to improve long-term retention (Doherty et al. Reference Doherty, Moulin and Smeaton2010) and to lead to higher performance on unrelated neuropsychological tasks compared to re-reading a personal diary (Silva et al. Reference Silva, Pinho, Macedo and Moulin2013). The ability to recall an event is also substantially improved (Hodges et al. Reference Hodges, Williams, Berry, Izadi, Srinivasan, Butler, Smyth, Kapur and Wood2006), but there are also some drawbacks, like an increase in false recognition and the risk of remembering another’s memories as one’s own (Silva et al. Reference Silva, Pinho, Macedo and Moulin2016).

While most studies involving SenseCam focus on patients with brain damage or memory impairments, some research has also explored its use among healthy individuals. These studies suggest that SenseCam may serve as a cognitive stimulant in everyday life (Silva et al. Reference Silva, Pinho, Macedo and Moulin2013). The goal of such research is not only to compensate for cognitive deficits but also to enhance cognitive functions. When enhancement is mentioned, it typically evokes images of invasive interventions, such as surgical procedures, brain implants, or drugs (Liao and Sandberg Reference Liao and Sandberg2008; Neubauer Reference Neubauer2021; Grinschgl and Neubauer Reference Grinschgl and Neubauer2022). From the standpoint of EXT, cognitive enhancement encompasses more than chips in the brain. The scope of the discussion should be broadened to include non-invasive cognitive technologies and artefacts (Heersmink Reference Heersmink2017), which are components of the human hybrid cognitive system. Enhancing cognition involves not only altering the brain but reconfiguring the broader cognitive ecology in which thinking takes place.

Activity/passivity

A key issue concerning memory devices lies in the degree of user involvement during encoding and retrieval processes. Transactive memory needs the input of the user: an effort is required to retain the location of the information. However, in the field of technological memory studies, it is common to underline the passivity of the user. SenseCam is a passive memory aid and requires no training or time to use. Such devices based on passive process relieve cognitive resources, whereas active process ‘requires executive skills (ability to plan, organize, problem-solve, and think flexibility)’ (Hodges et al. Reference Hodges, Williams, Berry, Izadi, Srinivasan, Butler, Smyth, Kapur and Wood2006, p. 190).

The difference is critical for people who have a brain injury and cognitive disorders. Active strategies, like keeping a diary, are time-consuming and cognitively demanding. The information encoded is subjective, and data are selected; it is impossible to record everything in a diary. In contrast, SenseCam represents a form of ‘total capture’ technology, passively and objectively recording continuous streams of experience. This feature has been shown to evoke stronger emotional responses than traditional diaries, and to generate a more vivid sense of reliving the past (Silva et al. Reference Silva, Pinho, Macedo and Moulin2013). Yet, this same passivity raises important ethical issues. Cognitive integration of technological resources risks undermining epistemic agency if users are no longer critically engaged with the resources they rely on. Meaningful cognitive integration should involve active engagement and the cultivation of epistemic virtues. The agent must not merely use technologies, but interact with them in ways that reflect intentionality, responsibility, and critical awareness (Menary Reference Menary2012).

Despite its well-documented benefits, the use of SenseCam also presents a number of practical limitations. In particular, issues related to device ergonomics and data visualization remain significant. The device can capture several thousand images per day (typically between 2,000 and 5,000), which raises substantial challenges for the organization, navigation, management, and retrieval of recorded experiences (Doherty et al., Reference Doherty, Moulin and Smeaton2010). To mitigate the risk of cognitive overload, some approaches advocate selectively recording salient events or clearly delimited episodes. However, as Doherty et al. point out, determining in advance which moments will prove meaningful is often difficult, which motivates more exhaustive recording strategies. Yet, such comprehensiveness comes at a cost, as locating specific memories within large image datasets can become laborious and tedious. In this respect, AI offers promising solutions, for instance by automatically segmenting events, identifying patterns, or clustering similar experiences. Further developments could aim to reduce the so-called ‘semantic gap’ by incorporating recognition technologies capable of providing contextual information about the content of images, such as the identity of individuals depicted. However, such approaches raise important ethical and legal concerns, particularly regarding privacy, image rights, and data protection (Teijlingen et al. Reference Teijlingen, Oudman and Postma2022). A sufficiently advanced version of SenseCam, capable of integrating temporal, spatial, and semantic information, could support the reconstruction of episodic memories in line with the what–when–where framework. In clinical contexts, such systems may even contribute to restoring forms of episodic recall that include second-order or reflexive contents. More broadly, these developments are in line with externalist approaches to cognition by illustrating how technological systems can become functionally integrated into cognitive processes, sometimes operating as genuine cognitive prostheses. As these technologies evolve, the boundary between biological cognition and technological artefacts becomes increasingly blurred, reinforcing the view that cognition is extended and hybrid.

Affective technology

The relationship between the technological environment and episodic memory also has implications for emotions and affective processes (Piredda Reference Piredda2020; Talarico Reference Talarico2022). Many people spontaneously document their lives on social media, without necessarily anticipating a future recall of these posts. Yet, many technologies have been specifically designed to reactivate these digital traces, stimulate nostalgia, or regulate emotions (Talarico Reference Talarico2022). While the voluntary recall of positive memories is a well-established strategy for emotional regulation, research shows that involuntary memories, often triggered by environmental or technological cues, have a stronger emotional impact than voluntarily evoked memories (Talarico Reference Talarico2022). This type of affective regulation is targeted by new technologies.

Environmental resources contribute to generating a whole range of affective states without action and intention from the user. For example, Google Photos often pops up memories and emotions by displaying ‘remember 2 years ago’. One can feel nostalgic or upset upon reviewing an unsolicited memory, like a weekend with an ex-partner (The ‘Spotify wrapped’ playlist also generates emotion via the music listened over the past year). Nostalgia can be considered an affective state associated with episodic memory. Unwanted past events might instead generate a negative affective state and unwanted memories that can be traumatic. To avoid unwanted memories, some platforms allow users to choose who is authorized to pop up spontaneously. The cognitive integration of such technology needs to find a good balance between automation and the activity of the user. Technological environments help individuals regulate cognition and emotion and are integrated into extended affective memory processes.

Extended phenomenology

Although EXT may seem counterintuitive, notably because we do not spontaneously experience the feeling of being ‘extended’, several phenomenological and ethical arguments reinforce its plausibility. A classic example is the cane used by blind people (Heersmink Reference Heersmink2017). The cane is not experienced as an object external to the person but as an extension of their own body and perceptual system. The cane is not part of the external world; it is what blind people use to encounter and explore the world.

Another example can be found in the psychological consequences of computer theft or hacking. Victims report feelings of dispossession, profound trauma, and even a form of rape (Button et al. Reference Button, Blackbourn, Sugiura, Shepherd, Kapend and Wang2021). This suggests that the technological environment, integrated into everyday practices, is experienced as an integral part of the extended self. The environment is no longer external but actively participates in cognitive processes and the constitution of the self. If our mental capacities can include artefacts properly integrated into our cognitive system, then any deliberate attack on these artefacts constitutes a personal attack in the same way as an assault directed against the body or mind (Carter and Palermos Reference Carter and Palermos2016).

Let us return to extended episodic memory in technological environments. Among the avenues for further research, one question arises: can we conceive of a truly extended episodic memory? Autonoetic consciousness, which is associated with the phenomenology of episodic memory, is the form of reflexive awareness through which the subject recognizes themselves as the agent who lived through the past experience. It is often described as responsible for the feeling of reliving the experience and has long been considered constitutive of episodic memory (Tulving Reference Tulving1985; Michaelian Reference Michaelian2016). However, recent studies challenge this idea by putting forward conceptual, methodological, and empirical arguments that undermine the need for autonoetic consciousness and a phenomenological component in episodic memory (De Brigard Reference De Brigard2024). Could it be possible to identify cases where episodic memory resides not solely in the mind of an individual, but in a hybrid system composed of a human subject and their technological environment, possibly understood as a cognitive prosthesis? This suggests that certain forms of episodic memory could emerge not at the level of the isolated subject but at that of an extended system in which the experience of reliving a past event is situated within the dynamic interaction between the agent and the technology.

Prospective memory

Prospective memory refers to remembering things to do; it is the memory of intentions, deferred intentions, and the ability to remember to perform an action at the appropriate time (McDaniel and Einstein Reference McDaniel and Einstein2007; Walter and Meier Reference Walter and Meier2014; Lecouvey et al. Reference Lecouvey, Gonneaud, Eustache and Desgranges2015; Pourjaberi et al. Reference Pourjaberi, Shirkavand and Ashoori2023). Prospective memory is typically divided into four forms depending on the type of cue that triggers recall of the intention: time-based, event-based, activity-based, and spontaneous recall, which arises in the absence of a cue. In time-based tasks, the intention must be recalled at a specific time or after a certain time interval (e.g., attending an appointment at 2 p.m. or taking medication after 45 minutes). Event-based tasks involve retrieving the intention upon the occurrence of a specific event in the environment, such as remembering to mail a letter while passing by a post office. Finally, activity-based tasks involve remembering to act before or after performing a particular activity, such as taking medication after dinner. Although this last category has strong ecological value, it remains little explored in the literature, due to its conceptual proximity to event-based tasks and the methodological difficulty of isolating it, particularly because it does not necessarily require interrupting an ongoing task. Experimental studies thus focus mainly on time-based and event-based tasks, in order to better understand the cognitive processes specific to or common to each (Walter and Meier Reference Walter and Meier2014; Lecouvey et al. Reference Lecouvey, Gonneaud, Eustache and Desgranges2015; Pourjaberi et al. Reference Pourjaberi, Shirkavand and Ashoori2023).

Unlike the retrospective forms of memory discussed above, which are oriented towards the past, prospective memory is future-oriented (I said previously that prospective memory was almost never studied in the philosophy of memory, but one aspect has been extensively studied, the fact of imagining the future, of mentally projecting oneself into the future (Lecouvey et al. Reference Lecouvey, Gonneaud, Eustache and Desgranges2015). Mental time travel has begun to be the subject of in-depth investigation (Perrin and Michaelian Reference Perrin, Michaelian, Bernecker and Michaelian2017) about episodic simulation and the relationship between memory and imagination.). It refers to the ability to form intentions about future actions and to remember to execute these intentions at the appropriate moment (McDaniel and Einstein Reference McDaniel and Einstein2007). In this sense, prospective memory concerns the planning of future actions as well as the retention and retrieval of intended actions over time (Walter and Meier Reference Walter and Meier2014; Lecouvey et al. Reference Lecouvey, Gonneaud, Eustache and Desgranges2015; Desaunay et al. Reference Desaunay, Postel, Bensaber, Gonneaud, Baleyte, Anger, Guénolé, Malvy, Bonnet-Brilhault, Eustache, Desgranges and Guillery-Girard2019; Pourjaberi et al. Reference Pourjaberi, Shirkavand and Ashoori2023). What is central for the present discussion is the ability to maintain future-oriented intentions and to retrieve them when required. Prospective memory is widely recognized as a complex, multi-component capacity that recruits several cognitive processes depending on task demands and context. It involves a declarative component related to the content of the intention, as well as executive control processes such as monitoring, attentional allocation, and cue detection (McDaniel and Einstein Reference McDaniel and Einstein2007). Metamemory processes may also influence prospective remembering by shaping strategy selection and confidence in one’s memory. Performance in prospective memory tasks may furthermore be influenced by subjective evaluations of the importance of the intended action. The perceived importance of an intention can depend on individual values, goals, and motivations (Walter and Meier Reference Walter and Meier2014; Lecouvey et al. Reference Lecouvey, Gonneaud, Eustache and Desgranges2015). In experimental settings, importance is often manipulated by the experimenter through incentives, rewards, or social motivations. While methodologically useful, such manipulations often lack ecological validity and fail to capture the everyday strategies individuals actually use to manage prospective intentions. This motivates a closer examination of real-world practices, including the externalization of prospective memory through technological resources (Gilbert et al. Reference Gilbert, Boldt, Sachdeva, Scarampi and Tsai2023).

A more thorough examination of the role of technological resources in prospective memory tasks could thus shed light on often underestimated compensatory capacities (Schnitzspahn et al. Reference Schnitzspahn, Ihle, Henry, Rendell and Kliegel2011; Annese et al. Reference Annese, Klaming, Haase Alasantro and Feinstein2023; Gilbert et al. Reference Gilbert, Boldt, Sachdeva, Scarampi and Tsai2023). From a perspective consistent with EXT, these strategies should not be understood as mere external aids, but rather as functional components of extended cognitive systems. Ecologically situated case studies illustrate how individuals sustainably integrate technological artefacts into their everyday memory practices.

Annese et al. (Reference Annese, Klaming, Haase Alasantro and Feinstein2023), for example, describe the case of AV, a patient observed both in laboratory settings and in his everyday environment. After an initial phase relying on rudimentary aids (such as handwritten notes), AV progressively incorporated digital tools into his prospective memory strategies. The use of computers, and later smartphones, is not limited to passive storage but actively contributes to the planning, maintenance, and execution of future intentions. Calendar and reminder applications provide regular notifications about upcoming events, support the initiation of actions at the appropriate moment, and also serve a reflective function, allowing AV to record personal notes comparable to a journal addressed to his ‘future self’.

In this type of configuration, the successful performance of prospective memory tasks no longer depends exclusively on internal processes but emerges from a hybrid system in which neural and technological resources are tightly coupled. This case, therefore, supports the idea that certain forms of prospective memory can legitimately be attributed to an extended agent–artefact system, rather than to the biological subject in isolation, thereby reinforcing the externalist approach defended by EXT.

Externalization of prospective memory

Beyond brain processes, the memorization of intentions also relies on the use of environmental resources. Individuals commonly use devices such as diaries, calendars, post-it notes, or even the help of others to maintain and update their intentions. This use of the environment, referred to as intention offloading, consists of externalizing a cue or reminder in the physical or social world to facilitate the future retrieval of an intention. Many daily practices are linked to our ability to transfer prospective memory to technological devices. For example, adding a niece’s birthday to a connected calendar ensures it comes to mind when a notification prompts us to send wishes. This extended mechanism thus allows the memory load to be distributed between the brain and the environment, increasing the efficiency of carrying out intentions in daily life. Prospective memory seems to be an excellent candidate for considering technological resources as an integral part of the memory system. Without extended resources, we will fail in most of our prospective tasks. A large study shows that ‘people rely more on external memory than internal memory for prospective tasks’ (Finley et al. Reference Finley, Naaz and Goh2018) and that ‘evidence demonstrates that external reminders are very effective’. Some experiments show that the forgetting rate decreases to 5 per cent instead of 45 per cent when participants use environmental reminders rather than brain memory for prospective tasks (Gilbert et al. Reference Gilbert, Boldt, Sachdeva, Scarampi and Tsai2023). Gilbert et al. confirm that extended mechanisms have received very little attention from psychology and neuroscience. Including environmental resources in prospective memory strategies offers valuable insights into memory functioning, with both theoretical and practical implications, including non-invasive treatments for memory disorders.

Prospective memory, health, and well-being

Prospective memory is essential because what matters in life often depends on planned tasks, like attending appointments or taking medication. Using a connected calendar and other environmental prospective memory strategies is a daily practice and has many practical implications, as it is effective and much more reliable than brain-based prospective memory. Redundancy and biloading are sound cognitive strategies, with external memory often appreciated for its dependable support in prospective tasks: ‘an alarm will never forget what time something is’ (Finley et al. Reference Finley, Naaz and Goh2018).

The ability to formulate, maintain, and execute intentions is a fundamental condition for leading an autonomous and socially integrated life. Intentions, as mental representations of future goals and actions, play a decisive causal role in the realization of our future projects (Bratman Reference Bratman1984; Pacherie and Haggard Reference Pacherie, Haggard, Sinnott-Armstrong and Nadel2010). Prospective memory is essential for daily functioning. It not only allows us to organize our time efficiently, but also to maintain a positive self-image and a reputation for reliability among others (Walter and Meier Reference Walter and Meier2014). An impairment of this faculty can harm self-esteem, autonomy, and even social integration. It constitutes a major challenge for daily functioning (Pourjaberi et al. Reference Pourjaberi, Shirkavand and Ashoori2023). By sustaining psychological well-being, prospective memory contributes to shaping the construction of a meaningful, structured, and relationally stable life.

Recent work in psychology highlights the importance of prospective memory for maintaining autonomy and well-being, particularly in individuals with cognitive impairments. A 2022 study (Scullin et al.) examined the effectiveness of smartphone-based strategies in supporting prospective memory, adopting a non-internalist approach that incorporated the technological environment as a cognitive resource. The results indicate that training in the use of these technologies significantly improves not only prospective memory but also the quality of life of participants. Two-thirds of those involved reported an improvement in their ability to remember to perform essential daily tasks, such as taking medication or attending appointments. These tasks, crucial for independent functioning, are particularly vulnerable in individuals with neurodegenerative disorders. In this sense, electronic aids appear to be one of the most effective interventions identified to date to support prospective memory. These findings support a broader conception of human memory, distributed between brain and digital artefacts, and demonstrate the positive impact of this perspective on individual cognitive health, autonomy, and well-being. Considering the role of the environment on cognition can have important implications for people with cognitive impairments. Smart technological devices, designed as cognitive prostheses that serve as memory aids, represent one of the most promising avenues for enabling people with cognitive impairments to manage daily tasks involving prospective memory.

As Scott & Gilbert (Reference Scott and Gilbert2024) point out, diary-based studies indicate that the majority of everyday medical failures involve not the recall of past events or semantic information, but the forgetting of intentions to be carried out (i.e., failures of prospective memory). This observation highlights a blind spot in contemporary philosophy of memory, which has largely focused on episodic and semantic memory, neglecting this particular type of future-oriented memory. Yet, it would be crucial for this field of research to fully embrace prospective memory, both to clarify its nature and to understand its philosophical and cognitive implications.

Metacognitive strategies

Metacognitive strategies play a central role in managing prospective memory tasks, particularly in the decision whether or not to externalize intention using devices. Memory offloading (i.e., delegating the task of remembering to an environmental resource) involves a cost–benefit assessment: the cost of setting up a reminder is offset by an increased likelihood of remembering the intention (Gilbert et al. Reference Gilbert, Boldt, Sachdeva, Scarampi and Tsai2023). However, this decision is not only a function of the objective difficulty of the task; it is also influenced by subjective beliefs about one’s own memory abilities. Thus, direct evidence of metacognitive influence on intention externalization would consist in showing that this externalization is predicted by subjective beliefs rather than objective memory performance (Gilbert et al. Reference Gilbert, Boldt, Sachdeva, Scarampi and Tsai2023). Notably, individuals often show a positive bias towards technological resources, using them more frequently than optimally necessary, which could be explained by a general tendency to avoid cognitive effort (Gilbert et al. Reference Gilbert, Boldt, Sachdeva, Scarampi and Tsai2023). This preference for reducing cognitive load has been widely documented in the literature, and use of technological and environmental resources is often preferred to purely internal management of prospective memory (Walter and Meier Reference Walter and Meier2014; Finley et al. Reference Finley, Naaz and Goh2018; Gilbert et al. Reference Gilbert, Boldt, Sachdeva, Scarampi and Tsai2023). Individuals strategically allocate their cognitive activity between brain and environment, based on criteria of efficiency and perceived effort.

To encode a prospective memory task, agents must identify appropriate cognitive and metacognitive strategies by choosing whether to remember them internally or to offload the task using environmental resources. This involves metacognitive assessment to monitor and control cognitive processes and abilities. Gilbert notes that offloading is predicted by metacognitive beliefs about memory capacities and by the agent’s confidence level. Thus, the intention to offload seems to be guided by lower confidence. One can also highlight the tendency to avoid cognitive effort and increase reliability. Setting a reminder requires time and energy, but it reduces cognitive load and improves the reliability of recalling the intended task. Studies show that awareness of metacognitive strategies is low and highlight the tendency to use environmental resources for prospective memory in a fairly automatic way (Finley et al. Reference Finley, Naaz and Goh2018; Gilbert et al. Reference Gilbert, Boldt, Sachdeva, Scarampi and Tsai2023). The use of technological resources for prospective tasks is an everyday and integrated practice. Information flow and agent activity/passivity are key elements of the extended memory framework.

Many everyday actions and intentions now unfold within digital environments. Online services are increasingly interconnected, forming ecosystems that can automatically manage information and future intentions on behalf of users. For example, when a user books a flight online, relevant details may be extracted from a confirmation email and inserted into a digital calendar, which then generates reminders without requiring deliberate input from the user. In such cases, the management of prospective intentions can be considered as delegated to the digital environment. To remember planned tasks, it is necessary to identify cognitive and metacognitive strategies. Prospective memory tasks involving technological resources are guided by metacognitive feelings, such as the feeling of having something to do or having an appointment scheduled. People often consult their calendars and rely on them when they experience this feeling. Given the widespread use of technological devices, retrieving planned tasks could also be an automatic process, without metacognitive feelings. This could be due to the habit of regularly consulting a paper calendar or the use of digital applications with contextual notifications (Future research could usefully investigate how social norms and cultural expectations shape preferences for internal versus external memory resources, particularly across different types of memory [e.g., prospective versus semantic]. Comparative and cross-cultural studies may help clarify the extent to which these preferences reflect efficiency considerations or socially valued conceptions of memory and cognitive competence).

Gilbert et al. (Reference Gilbert, Boldt, Sachdeva, Scarampi and Tsai2023) point out that refusing to use cognitive technologies, such as digital reminders, can be counterproductive, especially when they are easily accessible and proven to be effective. Empirical data indeed show that the forgetting rate drops drastically (from 45 per cent to 5 per cent) when using external reminders rather than memory alone. Thus, in a context of cognitive decline, such as in the elderly, refraining from using these cognitive prostheses in the name of principles such as ‘using one’s biological memory for fear of losing one’s mind’ could compromise the success of deferred intentions. This decision could lastingly influence prospective memory management strategies, to the detriment of autonomy and cognitive efficiency. The argument in favour of extended memory is strengthened: far from being a threat, cognitive technologies can play a key role in maintaining cognitive health and well-being.

From extended intention to EXT

An intention is a cognitive process and a mental state also involved in action. In philosophy of action, there is much debate about the status of intentions: is it a belief, a desire, or a distinct mental state? What can be said here is that prospective memory seems to be a particularly relevant candidate, perhaps the best, to support EXT. If we recognize that external traces are extended intentions, then we have a relevant and common case of an extended phenomenon. If intentions are mental states and they are extended, then it is not only cognitive processes that are extended but also the mind.

Intentions play a central role in human life, particularly in the planning and execution of future actions. According to Bratman (Reference Bratman1984), intentions are specific mental states, distinct from beliefs or desires, which organize action in a coherent and prospective manner. Unlike desires, they engage the individual in dynamics of organizing and coordinating plans and pursuing goals. In the context of prospective memory, it is future-directed intentions that capture our attention: they allow us to structure projects, prioritize goals, and organize daily life. Making a decision then amounts to forming an intention based on multiple motives, which implies an articulation between practical reason and cognitive structure.

Neuroscientific research on intentions and voluntary actions tends to focus on very brief timescales, which limits our understanding of intentional dynamics in broader contexts (Pacherie & Haggard Reference Pacherie, Haggard, Sinnott-Armstrong and Nadel2010; Lecouvey et al. Reference Lecouvey, Gonneaud, Eustache and Desgranges2015). Intention should be understood as a particular mental state, possibly correlated with specific brain states. Two core features define intentions: conscious accessibility and action-oriented, closely linked to future behaviour.

This dual dimension, reflexive and purposeful, distinguishes intentions from other mental states such as desires or beliefs. A quick analysis also reveals a diversity of modalities. Some intentions are fixed, such as medical appointments or professional obligations, entered into a calendar. Others, more flexible or potential, refer to options considered without a firm decision having been made: for example, noting in a diary several possible events for the same time slot, with a view to choosing later. These open intentions can be close to simple desires, without being devoid of structure or organizational function. Here again, metacognition comes into play to determine whether deferred intentions control behaviour more or less automatically, whether they are directly endorsed when the reminder arrives, or whether they are subject to re-examination.

Whereas Bratman (Reference Bratman1984) defends a distinction between intention and desire, emphasizing that with desire being the opposite of intention, it is possible, without logical contradiction, to want to play basketball while also hoping to finish an article on the same day. I argue that considering potential and multiple intentions is not inconsistent. This intentional flexibility demonstrates the richness of the phenomenon and underscores the need to study technological resources as cognitive extensions that contribute to the distributed management of intentions. Digital calendar, for example, can highlight the complexity of human intentions, their interaction with social context, and the difficulty of drawing a clear line between firm intention, uncertain planning, and simple desire.

It appears that, like much of the work on prospective memory, research on intentions and actions largely neglects the role of strategies involving the environment. Indeed, environmental resources and cognitive technologies are rarely taken into account in the analysis of the processes involved in the execution of prospective tasks. This omission constitutes a significant limitation, as it does not reflect the real conditions of human cognition, which is often distributed between the brain and the environment. Therefore, it becomes necessary to build bridges between the philosophy of action, the philosophy of memory, and EXT, in order to better understand how intentions are formulated, maintained, and realized in ecologically valid contexts.

Storing an intention in environmental memory relieves the brain of the need to retain the precise content of that intention (Gilbert et al. Reference Gilbert, Boldt, Sachdeva, Scarampi and Tsai2023). This externalization reduces activity in the medial rostral prefrontal cortex, which is involved in maintaining memory of the intention itself. However, activity in the lateral rostral prefrontal cortex, which is linked to action preparation, remains unchanged, as it is still necessary to remember that an action must be performed. The use of environmental resources produces differentiated effects on the brain regions involved in prospective memory: it alleviates the storage of the specific content of an intention while maintaining vigilance related to its execution.

Environmental resources store not only content, but also intentions, which are representations of future goals and actions. Calendars, diaries, lists, notes, and other personal information can be examples of what Gilbert calls ‘intention offloading’. Thus, it must be recognized that ‘intentions are stored in our extended physical and social environments’ (Gilbert et al. Reference Gilbert, Boldt, Sachdeva, Scarampi and Tsai2023). We rely on information from the environment to structure our lives. We carry our smartphones with a connected calendar everywhere. To remember an intention, we write it down (or the device automatically encodes it). When we need to remember the intention at the right time, we search for it, either intentionally or mechanically by looking at the calendar, or the appropriate memorized intention arises when we receive a calendar notification. We need our devices to carry out our prospective memory tasks. Applications such as Connected Calendar are technological resources that carry intentional actions (Inspired by Clark and Chalmers Reference Clark and Chalmers1998). Here we have a genuine case of extended intention offloading; we can claim that we and our environment constitute a united cognitive system.