Merged Memory - How Ubiquitous Information Can Affect Memory Formation

Abstract

Distributed memory research suggests that memories are not all located in the head but are spread across heterogeneous resources including bodily, social, and computational arrangements. As use of constantly accessible information resources such as smartphones and other Internet accessible devices increases worldwide dubbed here ubiquitous information resources (UIRs), the propensity to gain information becomes as easy as lifting a finger. For some this is a worry about the reliance on external systems over internal cognitive resources. This project examines through a distributed memory framework whether an increased use of external systems to supplement internal resources will have an affect on memory. Specifically, the project examines properties of UIRs in relation to people constant accessibility, limited sensory modality engagement, how these can affect metamemory and the consequent effects on what information one chooses to endorse. Finally, this project suggests that increased interaction with UIRs may replace existing social memory systems with technologically mediated transactive memory systems (TMTMs), changing the way in which both individual and collective memories form as this effect increases over time.

Introduction

Recent advances in cognitive science, philosophy, and psychology point to theories of cognition that emphasize the dynamic interactions between brain, body, and the environment, over more traditional computational approaches that purport cognition to be arising from a single entity, most commonly the brain. This view has been developed in different disciplines but varies within each one, and comes about under a variety of different guises. One area in which this theory has seen development over the last fifty years is memory research. From older psychological theories such as Vygotsky s theory of Mental Scaffolding (1987), Wegner s Transactive Memory System (TMS 1987 Wegner et al., 1985), sociological theories such as Halbwach s Collective Memory (Halbwachs and Coser, 1992), to more recent amendments in the cognitive sciences such as Clark and Chalmers Extended Mind (EM 1998 Clark, 2014, pp.192-211 Clark, 2008), and O Regan s Sensori-Motor Contingencies (Degenaar and O Regan, 2015), academics have used this dynamic, extended, and distributed framework to gain valuable insight into memory research.

These theories vary in their strength of application towards memory, and vary how memory is conceptualised. One prominent conceptual variance within each theory is where each theory places the memory. This varies from purely internal accounts, i.e. memories are only found in the brain, to theories that incorporate elements of the environment as a full extension of the mind external encodings were [ ] capable of becoming so deeply integrated into online strategies of reasoning as to share a great deal of the coarse functional role of internally tokened beliefs (Clark, 2014, pp. 193). Some, like Wegner s TMSs take a stance somewhere in the middle, suggesting that memories reside as distributed across groups, but only when the group is considered as a combined system (Wegner, 1987).

If we are to label purely externalist accounts, extended, and weaker accounts, distributed, evidence that is critical of the extended variation (see Adams and Aizawa, 2011) suggests that the distributed approach is a more accurate framework for memory research. For example, extended accounts such as Clark and Chalmers EM (1998) include strict criterion about when external resources can be labelled cognitive, such as automatic endorsement from the individual and previous conscious endorsement (Clark, 2014, pp. 197). Other critics suggest that theories of EM are associated purely with an epistemological concern about the nature of certain disciplines accepting extended frameworks would turn psychology into a kind of anthropology or sociology or ecology (Butler, 1998), but refuting it would keep psychology and the cognitive sciences in its individualistic roots (Rupert, 2009). This project takes Clark s (2014, pp. 208) proposal in that we should carve our cake to suit the varying experimental and theoretical purposes needed .

Hence, for the purposes of this project a distributed memory framework is used, one that does not commit to either side too strongly, positing memory as coupled with, situated among, or incorporating heterogeneous resources beyond the brain (Michaelian and Sutton, 2013, pp. 2), but not imposing strict criterion so that social and epistemological concerns are considered, whilst still benefitting from the computationalist methods that have been the traditional approach in memory research for the last twenty years.

There is a prima facie view in many academic disciplines that over time our dependence on material tools has increased. If we take the distributed memory approach in that memories incorporate heterogeneous resources beyond the brain (Michaelian and Sutton, 2013, pp. 2), this implies that over time our memories have become spread out over a vast array of items, from other neural to physical, and most recently computational arrangements. This story can be sketched out historically. Starting around 400BC, Plato argued that (Phaedrus (274-7)), the advent of writing diminished the mind as it led people to rely on external resources. His argument was that prior to writing most of the thoughts and concepts people had were not spread physically or computationally as they are now, but were kept in the head relying on written words diminished the content of the meaning and in doing so could damage the intellect. His statement can be characterized as a worry about mental atrophy as one s reliance on external resources increases, one s ability to form internal memories or thoughts is damaged or changed in some way. For the purposes of this project this can be called the atrophy principle.

A different but related worry was raised when the printing press was invented. Conrad Gessner worried that the amount of print being created caused an unmanageable flood of information (Bell, 2010) too much for any one individual to process. We can characterize this worry as the overload principle as there is too much information for any one individual to process, how does one choose which information to endorse? Ong (2013) exploits both of these principles by highlighting the ways in which new technologies have altered human thought and restructured consciousness (pp. 77), via an examination of the different medias and associated properties that each technology captures, for example writing forming context-free language or autonomous discourse [1] (Hirsch 1977, pp. 21 3, 26) when compared to oral speech. Applying these two principles to recent technological developments one could suggest that they have been highly exacerbated. The use of almost limitless Internet accessible informational resources, dubbed here ubiquitous informational resources (UIRs) allows for information at our fingertips (Sparrow et al., 2011), including the propensity to cognitively offload (Risko and Gilbert, 2016) to these resources. Recently some academics have seen this is a risk suggesting that varying the interactions we have with external symbol systems can dramatically alter the capacity and operation of human memory (Michealian and Sutton, 2013 Donald 1991: 308-333 Donald, 1998).

Hence, by a critical analysis of empirical research that evaluates agent-technology interactions via a distributed memory framework, this project assesses the impacts an increasing use of UIRs will alter the way in which humans form and share memories, suggesting cognitive explanations as to why these impacts occur.

To start, I will begin by sketching an outline of memory using both distributed and functionalist approaches, highlighting both the empirical and theoretical components that a distributed framework will favour in relation to this project.

Functionalist approaches

As mentioned in the introduction, theories of mind in cognitive science have traditionally followed a computational model, in that aspects of mental function, such as memory, are broken down into their functional roles (Fodor, 1981). This functionalist perspective has helped define and categorize standard conceptions of memory, although recently this has seen its drawbacks (Thelen and Smith, 1994). The most common functionalist account of memory splits it into three distinct yet overlapping processes: encoding, consolidation, and retrieval (Winters et al., 2008). Encoding is the conversion of environmental stimuli into neural activity, for example the conversion of light signals from the retina of the eye into action potentials in the brain (Treisman Kanwisher, 1998). Consolidation is the process whereby memory becomes more integrated within the brain, typically seen as the movement of memory through three stages: working-memory to short-term memory, to long-term memory[2] (Baddeley and Hitch, 1974). Retrieval under this model is the movement of memory trace from long or short-term into working memory. Why and when retrieval occurs has been shown to depend on a vast variety of factors, such as context (Eich, 1980), attention (Craik et al., 2000 Dudukovic et al. 2009), state-dependence (Turner and Engle, 1989), and interference (Cohen, 1989).

The functionalist model of memory outlined above treats external information, as at best cues or triggers to the real memory processes going on in the head (Michaelian and Sutton, 2013, pp. 2), however philosophically this has been shown to create problems such as the Cartesian Theatre[3].

Nevertheless, the stating of this philosophical position need not take away from the use some computational concepts have with our understanding of memory. For example, one can still use the differential memory types that have come from empirical research in cognitive psychology whilst using a distributed memory framework. Hence, this project utilises a distinction between internal and external memory to try and further elucidate their relationship: Internal memories as those types not able (at current) to be spread externally, whereas external memories, such as photos, or print, are those that can be spread across other mediums besides the neural medium. To explicate this and gain a better understanding of the ways in which internal memory may be affected as we increase our interactions with UIRs, a categorization and analysis of the main memory types to be affected by the change in information accessibility is posited.

Similarly in addition to functionalist methods defining how temporality is a constituting factor in memory storage, it has also done a lot to categorize different types with some neurophysiological evidence to back these models. Modularity suggests that there is mapping between functional specialization and specific brain areas however this dichotomous view has been suggested to oversimplify the matter (Pessoa, 2008). Neurophysiology aside, three classes of memory type relevant to this project exists have been theorised: declarative, episodic, and autobiographical. Declarative memory distinguishes itself from episodic memory by relating to only the structured knowledge of an experience whereas episodic incorporates more: the mental experience of re-experiencing that accompanies the products of the constructive synergy of memorial engrams (Perrin and Rousset, 2014 Tulving, 1972 1985). For example someone visiting Sussex University may be able to remember the name of the square he visited Library Square the declarative memory, but may not be able to remember the smells of the bushes or how windy it was that day. These extra pieces of memory that can t be structured are known as the episodic memory.

Both declarative and episodic memory makes up autobiographical memory but only in reference to one s life. So, the memory of the lead actor in a film, such would not be autobiographical unless there was an associated story that accompanied the knowledge of meeting the said actor. As use of UIRs increases suggestions are made that our episodic autobiographical memory is being converted to semantic memory, semantic memory acting as a heuristic for episodic content (Fawns, 2011).

Empirical memory research on agent-technology interactions

Many theories exist as to how recent technological developments may be altering human thought, and the way we process and remember information (Rowlands et al., 2008 Ong, 2013). However, only a limited selection exists that is memory-based empirical research on agent-technology interactions (Sparrow et al., 2011 Barr et al., 2015), relevant to this project. I will summarize one piece of empirical research on memory function and agent-technology interactions, in addition to the consequences it suggests will occur as we increase our use of UIRs.

Sparrow et al. (2011) studied memory function during agent-technology interactions. The aim was to find whether the Internet had become an external memory system that is primed by the need to acquire information. The authors ran four experiments as part of the study, all with different implications, that seemed to suggest the above hypothesis, three of them being relevant for the purposes of this study. The first implication was that the computer may be primed when the concept of knowledge in general is activated (Sparrow et al., 2011, pp. 1). This implication came from a colour stroop-naming task that compared interference from computer-related words, such as laptop , and other non-computer related terms. More interference occurred from the computer-related terms implying the statement above.

The second implication was based around directed forgetting (Bjork, 1972). Participants were given 40 trivia statements to remember in the style of information they might look up online. They then typed them into a computer to direct attention in addition to allowing for a more generous test of memory. Half of the participants were told the statements were going to be erased whilst the other half were told that the computer would save what was typed. After being shown and typing the statements participants were told to remember as many as they could. A significant effect found that more statements were remembered from the group that believed the computer would erase what was typed.

The final implication from the study showed that individuals were better at remembering where , not what in relation to information stored on a computer. Participants were given 30 generic statements to remember, such as An ostrich s eye is bigger than its brain , and were given a generic folder name where they had come from such as FACTS . The participants were then told to write down as many statements as they could remember. After writing down as many statements as the participants could remember the participants were given a cue to one of the statements and asked which folder it had come from, for example for the ostrich statement above, the cue being What folder did the statement about the ostrich come from? Figure 1 shows that the participants were much better at remembering from what folder the statement came from over what the statement said. The authors conclude this to be unexpected given the memorable nature of the statements however certain criticisms need to be voiced here. First, the folder names were small and capitalized, possibly leading to saliency bias (Peleg et al., 2001), second, although the trivia statements were posited by the authors to be more memorable, their length could have made them harder to remember, in accordance with studies of short-term memory bandwidth (Magnussen et al., 1991).

These three experiments imply that as use of UIRs increases the way in which we encode and process our memories is changing. One way in which the findings of this research can be understood in terms of action, is how using UIRs can have an impact on what information one chooses to endorse, in addition to how individuals select between internal and external resources.

The selection problem

Ideas about how one decides what information to endorse and what information to select were first hypothesized thirty years ago (Lovett and Anderson, 1996 Lovett and Schunn 1999 Payne et al., 1988 Reder, 1987). Cognitive psychology supposes that before answering a question or solving a problem, mental strategies about where the information came from varies. For example, mental strategies for multiplying vary via two cognitive paths. For example one can either use mental multiplication, or one can take retrieve the answer from one s lexicon. Under a distributed memory framework this problem extends what processes govern an agents choice between internal mental strategies or external resources? Or in other words, if asked a question, how does one know when to use UIRs over their own reasoning. Usually this choice is easy dependent on the type and specificity of the question asked. For example, general knowledge questions and answers, such as the birthdate of a famous person that is unlikely to be known by an individual, the individual would more likely resort to the external resource. Similarly, questions about mathematics can either be cognitively offloaded via tools such as a calculator to reduce the cognitive load, or solved internally (Walsh and Anderson, 2009), Similarly, as expressed in the EM (Clark and Chalmers, 1998), mental manipulation problems can be solved either internally or externally via props and tools. After selecting for information either internally or externally the agent then has to decide whether to endorse the information or not to form a belief (Arango-Mu oz, 2013 pp. 140).

The endorsement problem

This stage has been denied any functional role by some (Bernecker, 2010), assuming that information retrieved internally is already endorsed, however it does have empirical backing from cognitive psychology that shows not all information one retrieves internally is endorsed (Hertwig et al., 2008). Note that this problem does not work with Clark and Chalmers EM framework, as with the case of Otto one criterion of the EM is that information be automatically endorsed (Clark and Chalmers, 1998 Clark, 2014 pp. 197).

The selection and endorsement of either internal or external information depends on various factors. For example, external strategies cost more to the sensory-motor system, and usually take longer (Kalnikait and Whittaker, 2007), whereas internal strategies have more cognitive costs, including working memory and attentional costs. These factors must be estimated by the individual on a sub-personal level for the individual to be able to make quick and reliable decisions. Arango-Mu oz (2013) posits that our reliance on external memories, in particular those formed from agent-technology interactions, is managed via one s metacognition. He suggests that metacognition produces meta-cognitive feelings: phenomenal feelings that are produced by sub-personal monitoring and control of cognitive processes (Mendes et al., 2008 Dokic 2012 Proust 2010). These feelings work as they allow the agent to make decisions in an adaptive manner and are reliable (Sparrow et al., 2011 Reber Unkelbach, 2010).

Metamemory

Metamemory has been used to refer to both one s Theory of Mind (ToM), and to self-control. For the purposes of this project I will use metamemory only for its self-control reference. Metamemory works by monitoring and controlling cognitive processes, such as the endorsement of information and works via heuristics and cues (Nelson and Narens, 1990). These heuristics are estimated by the production of phenomenal experiences or feelings that point towards mental dispositions of the subject such as knowledge, ignorance, or uncertainty. Examples of these are feelings of knowing (Reder, 1987, 1996), forgetting (Halamish et al., 2011), error (Arango-Mu oz et al., 2013), and confidence (Koriat, 2008 Brewer and Sampaio, 2012). Metamemory works within the distributed framework as it is a dynamic component in the coupling of agent and environment, and not just a control mechanism regulating Cartesian mental performance (Kirsh, 2005). This can be explained in relation to selection and endorsement problems, in that sub-personal monitoring that is sensitive to heuristic value actively engages the individual in either action with an external object, or endorsement of an internal piece of information, Arango-Mu oz (2013) suggesting that these metacognitive processes provide awareness to the individual about when to consult one s external or internal resources.

The interactive hypothesis of metamemory posits that if an agent is not familiar with a piece of information then how accessible the piece of information is will affect whether one endorses it (Koriat and Sadot, 2001). It is a combination of both the cue-familiarity (Miner and Reder, 1994) and the accessibility hypotheses (Koriat and Goldsmith, 1996). Quicker responses tend to occur when the agent is familiar with the cues with slower responses when the agent is unfamiliar using both how accessible the information is in addition to cue familiarity. Again, under a distributed memory framework this can apply to both internal and external memories. Hence, accessibility and relevance are both heuristics that act on sub-monitoring which can lead to confidence in a specific memory. Similarly, accuracy and efficiency determine whether an external memory device is used (Kalnikait and Whittaker, 2007). The below figures portray two examples of situations in which metacognitive processes regulate decision-making within an agent and can be disrupted by the use of UIRs

Figure 2a) represents what happens when an agent has to answer a question on his or her own. After being asked a question metacognitive processes stimulate phenomenal feelings which either give positive responses, feelings of knowing (FoK) if the agent feels they know the answer, or negative responses, feelings of error (FoE), if the agent does not know the answer. Figure 2b) represents a conceptual example of what may happen when UITs are present. UITs may affect the agent after Sparrow et al. s (2011) suggestion that they we are forming TMPs with our technologies, in that one would assign expertise to their UIT (Peltokorpi, 2008), thus reducing the impetus to both cognitively endorse and select internal resources, decreasing the feeling of knowing (FoK), and the likelihood of being able to answer the question on their own. Similarly having access to a UIR could override the metacognitive processes involved in accessibility. As the information on the UIR is constantly accessible by the agent, after use with a UIR the likelihood of using his or her own internal resources decreases.

The above examples illustrate how, as use of UIRs increases worldwide, the impetus to use an external resource far outweighs the accessibility of drawing on one s cognitive resources. This automatic inclination to use external devices over internal cognitive resources can be damaging to the encoding processes involved in the agent.

Damage to encoding processes

Psychological studies suggest that this automatic inclination to choose an external resource over an internal resource can be damaging to encoding processes. Firstly, encoding is disrupted because knowing one has the option to retrieve information from external resources means there is less likely to be an intent to remember, which intuitively and empirically has been shown to improve encoding processes (e.g. Block, 2009 Buckner et al., 2001). Secondly, information accessed online is less likely to be emotionally integrated (Labar and Cabeza, 2006). As information accessed on a device is accessed (at current) via only visual and auditory sensory modalities, there is less likely to be emotional integration when consolidating the memory[4] (Herz, 1998). Lesion studies suggest that emotions mediate memories via the amygdala giving greater long-term memory for emotional than neutral events (Labar and Cabeza, 2006 pp. 56), however it should be noted that some scholars suggest emotions and cognition should be strongly non-modular (for a review see Pessoa, 2008), with emotions and cognition being realised over multiple brain modules.

On a more theoretical basis Fawns (2013) complements this view by suggesting that an increased reviewing of digital photos online, can lead to an avoidance of episodic memory. His work follows the view that semantic memory can act as a heuristic device for episodic memory (Tulving, 1986 pp. 400 Tulving et al. 1983 pp. 400 Wheeler et al., 1997 pp. 349), in doing so increasing semantic content but decreasing emotional integration. For example, imagine an individual who spent half of his life online, reviewing digital photos of past experiences he had had. When pondering a memory, what may first be re-collected would less likely be the original experience but a memory of the experience that he had stored online, in doing so the re-collection attributing a decreased number of elements of the original experience, including the emotional content.

Discussion

The aims of the present study were to find out whether living in an age that implements UIRs would have a negative impact on the ability to form memories, in doing so highlighting the empirical components a distributed memory approach favours. Looking specifically at i) a critical analysis of empirical research conducted on agent-technology interactions and ii) the effect on metamemory and consequent effects on action.

Distributed memory posits that memories can be distributed across items in an agent s environment, including bodily (Clark, 2008), environmentally (Hutchins, 1995a) and socially (Wegner et al., 1991) configured circuits. Memories distributed socially have perhaps intuitively seen more research from the social sciences (e.g. Halbwachs and Coser, 1992), than from the cognitive sciences or psychology. An alternative interpretation as to the results of the empirical research that would further explain the consequent implications on metamemory is that as we increase our interactions with technology, memory systems between social groups are being replaced by memory systems between agents and technology (Sparrow et al., 2011). To explicate this we present a foray into how memories might be distributed socially from a psychologist s perspective: Wegner s, Transactive Memory Systems (TMSs Wegner et al., 1991).

Transactive Memory Systems

TMSs are non-reductive systems that have distributed properties whereby individual memory systems [ ] become involved in larger, organized social memory systems that have emergent group mind properties not traceable to the individuals[5] (Wegner et al., 1991, pp.923). First formulated after studying couples in close relationships, Wegner et al. (1991) tested the hypothesis that couples in close relationships use TMSs. A comparative memory retrieval task tested groups of close couples, and groups of two non-acquainted individuals to remember items from different categories of questions as a couple. Individuals within the couple were either given explicit assigned categories to answer from or were given no explicit instruction. The author s suggested that as the close couple would already have implicit memory organization, they would perform worse on the memory retrieval task when given assigned categories, but better when given no instruction. The new system of categorization would interfere with the already implicit TMS in doing so interfering with retrieval processes, leading to better retrieval for the non-acquainted couple assigned categories than the close couple assigned categories. The results proved significant (see Figure 2) with 31.40 items retrieved on average when no assignment was given as opposed to 23.75 items when categories were assigned, for the natural couples, and 30.14 for assigned categories against 27.64 on average for the non-acquainted couples.

Wegner et al., (1991) originally conceived TMSs as implicit memory structures between two individuals. They have since gone on to postulate that TMSs can exist between groups (Peltokorpi, 2008), although no clear, explicit distinction is made between different levels of analysis in TM theory (Ibid., 378). TMSs occur because people implicitly learn the expertise of the other groups members. Could it be that TMSs between people are being replaced as TMSs between agents and technology? Although not explicitly discussed as a TMS, it has been suggested by several scholars that memories can also be distributed between items in one s environment, some scholars suggesting that it is precisely these mnemonics or exograms (Donald, 2010) in one s environment that allow for humans to both complete tasks that were previously impossible, in addition to surviving in a culture that exhibits mental overload . For example, Hutchins (1995a) discusses that the memory of how to fly a plane only emerges, if the pilot is in the presence of the cockpit, the speed cards , and the co-pilot of the plane The memory is not contained in any one of them, but emerges through their well-coordinated interactions (Fagin et al., 2013).

Technologically Mediated Transactive Memory (TMTM)

The implications from Sparrow et al. s (2011) study suggest that transactive memory with one s technology dubbed here technologically mediated transactive memory (TMTM) is replacing traditional TMSs between social groups. These suggestions were made as the behaviour the participants exhibited reflects how one might behave when in a social TMS. Participants rapidly thinking of computers when they find they needed knowledge , forgetting items that they believed would be remembered by the other partner, being primed to remember items they believed the other partner would forget, and being better at remembering where an item is stored over the identity of the item itself (Ibid., pp. 778). This move from social TMSs to TMTMs could have negative consequences for the individual outlined below.

Disadvantages of TMTM

Firstly, replacing social TMSs with TMTM is intuitively going to be negative, as one s interactions with the Internet do not promote sociality. Many scholars (see Fagin et al., 2013, pp.95 Arango-Mu oz, 2013) argue that it is precisely this promotion of sociability that gives collective memory its adaptive value and in turn provides a foundation for collective identity . Fagin et al., (2013) suggests that memory is constituted so that [ ] initially disparate renderings of the past become more similar (Fagin et al., 2013, pp. 95). Their work argues that memories flaws, i.e. its transient nature, its unreliability, and its malleability have adaptive value (Newman and Lindsay, 2009 Schacter et al., 2011 Bjork, 1989). When people converse in social transactive memory systems, social contagion , whereby memories suggested by another person are believed to be one s own (Fagin et al., 2013 pp. 95) and socially shared retrieval-induced forgetting occur where although people might be hearing incorrect information, these processes in turn promote mnemonic convergence. Mnemonic convergence allows for people to form a collective memory of the past that has in sociological literature (Halbwachs and Coser, 1992) been shown to provide a basis for collective identity.

Although one can socialise via technologically mediated means, and thus promote sociality in a certain sense, the rigidity and permanence of information stored online disallow for processes such as mnemonic convergence to occur, thus delineating from the formation of collective memory and collective identity. This follows the hypotheses that the constant reviewing of autobiographical content online can be damaging for identity development (Ward, 2013 Fawns, 2013).

Secondly, by the nature of the devices we use to access information online we can only access information via visual or auditory modalities, and increasingly the tactile modality (Obrist et al., 2016) amongst others[6], leading to a weaker consolidation of the memory. Increased repetition of technological means to consolidate memories has been suggested to allow for semantic knowledge to act as a heuristic for episodic knowledge (Fawns, 2013), which can lead to less selectivity when endorsing knowledge creating a weaker sense of self (Conway, 2005). Research on how emotions are constituted from perceptual experience, in addition to their having an impact on how long a memory is consolidated for complements this view.

Although both these points provide a case for social transactive memory systems over technologically mediated systems, it needs to be noted that there are negative consequences to socially distributed memory.

Advantages of TMTM

Social contagion and collaborative inhibition (Fagin et al., 2013) occur during social remembering, which although help to promote mnemonic convergence and sociality in turn helping form collective memory, however it is noted that in certain cases it be more positive for the society and individual to lose these processes than the adaptive value of collective memory.

As we share things online there are similar analogues to the processes that promote collective memory on social media in the form of algorithms. This is not surprising when social media is in essence trying to emulate social groups, so again could be seen as advantage of TMTM. Algorithms dependent on individual users cookies control what information one sees when one browses social media platforms. These algorithms are attuned to the information one has already seen before to provide information that is most relevant to the individual. As before memory is constituted so that [ ] initially disparate renderings of the past become more similar , online memory is also constituted so that initially disparate renderings of the past become more similar (Fagin et al., 2013, pp. 95) the difference being that one is constituted via algorithmic methods whereas the other is constituted via socialising. This top-down control of how collective memories are constituted is worrying for human agency (Sparrow and Chatman, 2013).

Using TMTM one can spread oneself too thinly across too many cues to memory changing the balance of memory between rich experience to surface knowledge of ourselves (Fawns, 2013), in addition to losing agency over the way one forms collective memory.

Future Research

The present research suggests that use of UIRs can affect metamemory and consequent endorsement of one s internal resources, however does little to suggest differences in this effect. A study to discern differences in this effect could be conducted, including both general questions to gain a picture of general use, like What age did you gain your UIR? How long have you used it for? and How often do you use it for per day? in addition to questions relating to metamemory like, When faced with a question how likely is it that you will resort to a UIR to answer it? This could discern the participants into two groups, those that use their UIR regularly and those that do not. The two groups could vary in their level of device interaction defined by high parameters on the questions above such as, gained device at age 12, used for 6 years now , over low parameters such as, gained first device at age 15, latest device had for 6 months . The hypothesis would follow that those who had a high level of interaction would be more automatically inclined to use their device before consulting their internal resources.

To clarify and validate the effects that using TMTM over social TMS will have on memory, in addition to furthering research on the effects maintained consolidation of memories from limited sensory modalities could have, a study that compared individuals who reviewed information online, for example over two hours per day, often over those that reviewed information less often, for example under 3 hours per week. If semantic knowledge is acting as a heuristic for episodic content (Fawns, 2013) for the more often group it is suggested that their memories may be less emotionally mediated. Functional Magnetic Resonance Imaging (fMRI) studies comparing activation in the amygdala during memory retrieval across groups, with less activation for the more often group would give weight to this hypothesis. Additionally, verbal reports that compare responses so a quantitative scale of the memory people most remember, for example on a scale of 1-10, 1 being the least, 10 being the most how much do you remember the different aspects of a given memory from last week?

Michaelian and Sutton (2013, pp. 18-19) propose possible directions that distributed memory approaches can go in terms of their metaphysical and mechanistic status including: How agent-technology-agent couplings work, the implications of human communication research on distributed memory research, what implications distributed memory has for the ontology of group cognition, and what cognitive mechanisms mediate the interactions between biological and external memory. Although the first three have not been answered during this project, we hope the metacognitive framework proposed in this paper has gone some way in answering the last question. An altercation of the study hypothesized above could allow for more empirical validity being attributed to this metacognitive framework and explicate which hypothesis of metacognition is most empirically sound so that one can use this during future studies.

Conclusion

As Bell (2010, pp.1) points out, worries about information overload are nearly as old as information itself . This project reviewed empirical research on agent-technology interactions finding that as UIRs increase people become inclined to search for information over internally re-calling it. This effect was then theorised to occur via processes involved in metamemory, in addition to suggestions that this may be down to a replacement of existing memory systems between social groups, to newly formed memory systems between individuals and technology. Although these interpretations have theoretical validity, future research will need to be carried out before the empirical validity of these results can be confirmed.