Memory is essential for all processes, as errors occur when necessary knowledge is inaccessible. While memory capacity varies among individuals due to factors like motivation, expertise, age, and cognitive abilities, design should prioritize the level of attention each individual can or will provide rather than solely focusing on memory capacity differences. This review focuses on the convergence of top-down and bottom-up processing within awareness, particularly in the context of working memory. It explores how the concept of cognitive load, along with its limited attentional resources, interacts within this framework. Additionally, it delves into the relationships between time constraints and the volatility nature of the system. The review also takes into account emotional factors such as motivation and anxiety. It proceeds to present a design case that examines how cognitive load under a stressful situation impacts working memory and decision making, particularly on non-intuitive application such as Trust Wallet—a cryptocurrency wallet that allows users to store, send, receive, and exchange various cryptocurrencies and tokens.

Memory & Cognitive Load

Information flow and memory interaction is a multi-step process starting with the transfer of information from the visual cortex into sensory memory buffers, and then into short-term working memory which moderates attention, decision making, and interactions with long-term memory. Long-term memory interactions include assimilation, accomodation, and transversal of semantic network frames to retrieve information, scripts, and mental models (Baddeley, 2012). Failures can occur at each step: (1) effective pattern detection will not occur in sensory memory if just noticeable difference thresholds are not met then, (2) excessive load and time constraints impact working memory, (3) while experts can quickly access tuned resources, shallow long-term memories might not be readily available for non-experts (Sweller, 2010). Stanford Psychologists Richard Atkinson and Richard Shiffrin proposed a memory model consisting of a sensory register along with memory store for both short-term and long-term (Atkinson & Shiffrin, 1968). Noting some inherent shortcomings in the Atkinson-Shiffrin model, Baddeley and Hitch proposed the concept of a limited-capacity short-term memory store consisting of a visual-spatial sketch pad, a phonological loop that stores verbal inputs and allows for rehearsals, and a central executive for attentional control (Baddeley, 2012 & Baddeley, et al., 1974). An episodic buffer was added to accomodate interactions between working and long-term memory (Coolidge, et al., 2005). Proactive or retroactive interference created by non-agreement between new and long-term memories cause errors (Nieuwenstein, et al., 2014 & Shipstead, et al., 2012). Anxiety, fatigue, and alcohol abuse add load on working memory (Staal, 2004). Sweller noted three categories of cognitive load that should be considered in product design: intrinsic which is the minimum to allow system interaction, germane associated with the assimilation and accomodation of knowledge and requires effective management, and extraneous which should be eliminated (Sweller, 2010).

Working Memory is Limited Capacity

While the Baddeley and Hitch model explains the functional aspects of working memory, it does not account for its capacity limits. Capacity models of working memory consider it holding a small number of discrete objects, typical limited to Cowan's four, Miller’s magic number 7 ± 2, or with more recent limiting theories setting a maximum of three or four slots (Cowan, et al., 2007 & Miller, 1956). The slot model is the most definitive in that items either get a slot and get remembered, or do not get remembered at all (Wei, et al., 2014). The concept of slots links directly to the notion of spatial binding or chunks, and chunks can be mapped into the idea of long-term memory schemas. Sweller described this end-to-end relationship as adding to the overall capacity and efficiency of working memory, and enhancing rehearsal functionality (Sweller, 2003). To account for the high efficiency of experts, psychologists Anders Ericsson and Walter Kintsch proposed a theory of long-term working memory where part of long-term memory works as a seamless extension of working memory to extend its capacity. Cowan believed experts were holding larger and more complex pieces of information in slots and implementing tuning to improve efficiency (Cowan, 2014 & Ericsson, et al., 1995). Items existing in working memory decay as attention is switched away (Barrouillet, 2004). Research suggests similar items in working memory, or signals across different modalities, can interfere with each other (Cowan, 2014). Stress, anxiety, and fear can create interference problems and reduce available capacity as well (Staal, 2004). Resource capacity is further limited by the need for attentional resources (Cowan, 2010).

Working Memory is Time Constrained

Time-based working memory models generally set a limit of 20-30 seconds until information is significantly degraded or gone. Psychologist John Brown put forth a theory of forgetfullness based on the concept that memory has traces which decay and become deactivated as time passes (Brown, 1958). Cowan's activation model proposes that activated memories begin to fade between ten to twenty seconds without rehearsal, and rehearsal can be more effective if information is chunked (Cowan, et al., 2010). Visuospatial information on an object (what an object is and where the object is located) can be maintained longer and refreshed via rehearsal (Coolidge, et al., 2005). Baddeley & Hitch suggested about two seconds of speech could be retained in working memory via silent rehearsal (Baddeley & Hitch, 1974). The time-based resource-sharing model (TBRS) states that problem solving in working memory requires both temporary storage and computational power, and that these are both limited by attentional resources (Barrouillet & Camos, 2012). The TBRS model reinforces the need for the sustained executive attention proposed by Engle (Shipstead & Engle, 2012). Information decay will also be exacerbated when multiple tasks require the same limited visual/spatial sketchpad, phonological loop, or attention (Cowan, 2016).

Working Memory is Highly Volatile

Multiple items held in working memory are not all held with the same strength or accuracy. The recency effect describes how newer items will be easier to locate, while earlier items may be decayed or forgotten. Oxford psychologists Sanjay Manohar and Masud Husain showed that adding irrelevant features to existing memory sequences caused information degradation, loss of attention, and performance declines via disruption and interference (Manohar, et al., 2016). Attention and working memory can be affected by distractions, as well as by proactive interference (Shipstead, et al., 2012). New data can either enhance tuning or create retroactive interference (corrupt) (Niewenstein & Wyble, 2013). Interference is particularly noticeable when new memories have similarities to existing memories (Underwood & Postman, 1960). Volatility is enhanced with age as required attention processes are more easily impeded from interruptions (Adams, et al., 2022).

Emotional Subsystems

Threat of danger has a powerful emotional impact on human cognition. Baddeley modeled the central executive on Norman's Supervisory Attentional System (SAS) which postulates that existing working memory tasks will be overriden when danger or threat is perceived. Kane and Engle verified experimentally that this has a neurological basis as well. (Kane, et al., 2002 & Coolidge, et al., 2005). Human perception is so attuned to threat detection that even non-existant dangers can be anticipated or believed to be dangerous. This evolutionary principle of threat detection links into Elika Razmjou's arousal theory which states that arousal regulates and initiates stress response in humans (Staal, 2004). Motivation and anxiety are forms of arousal necessary for humans to take action, and need to be considered in user-centric design (Norman, 2004).

Motivation

Psychologists Robert Yerkes and John Dodson created the Yerkes-Dodson law (inverted-u model) which modeled the relationship between stimulus (pressure) and motivation. Between minimally stimulutating tasks and overwhelming stress where performance decreases, exists a level of stimulus that will produce optimal motivation in humans (Staal, 2004 & Hattie, et al., 2020). Multiple theories expanded on this to include up to twenty different motivational items, and psychologist Sandra Graham noted the commonality across all theories was the concept that motivation "is the study of why humans behave the way they do, or why we do this instead of that" (Hattie, et al., 2020). Psychologist Mihaly Csikszentmihalyi used the term "flow" or "flow state" to describe when humans are motivated to achieve optimal performance, either intrinsically because of enjoyment, or extrinsically for personal reward (Csikszentmihalyi, 2005).

Anxiety

Anxiety can have a positive influence on performance if it does not extend to the point of producing overwhelming negative thoughts or extensive fear (Staal, 2004). Internalized anxiety can be produced by low self-esteem, the belief of reward loss, or the drive to perfectionism. External sources can be threats, fear of injury, or environmental (Csikszentmihalyi, 2005 & Norman, 1988 & Staal 2004). Psychologists Michael Eysenck and Nazanin Derakhshan developed the theories of processing efficiency, and attentional control, to describe how states of anxiousness affect cognitive performance, attentional ability, and goal completion. These theories note that limited levels of anxiety can lead to better effort, increased performance, and goal achievement (Derakshan, et al., 2009). Workload stressors created by time pressure, constant task switching, or environmental conditions create anxiety, degrade memory quality, and diminish cognitive performance (Staal, 2004). Trust creates credibility levels which reduce load and enhance performance, while fear creates addtional load (Norman, 1988 & Csikszentmihalyi, et al., 2005).

A User Experience on Trust Wallet

At a very late hour in June 2021, a novice crypto enthusiast received a message from a friend to check out a new token, Football Stars (FTS). The friend was pleased to share a screenshot of his original $15 investment that grew to be worth over $5,000 within a couple of hours. He then encouraged the user to purchase some for herself via Trust Wallet which was among the few FTS supported systems at the time. The user would normally perform research before making any purchase decision, however, she decided to trust her friend to have gone through the process and started to transfer Binance coins (BNB) from Binance to Trust Wallet so she could swap for FTS by connecting to PancakeSwap, a decentralized exchange (DEX). Immediately, Trust Wallet showed her initial $200 investment of 650,000 FTS with a value of over $70,000 in balance. Highly motivated, she decided to purchase another $1,000 for a total of 32,291,161 FTS tokens that increased the balance to over $400,000 in value. The user's excitement reached an all-time high, but her joy was short-lived when she saw the $400,000 balance began to drop after a few hours. In the state of panic, she tried to trace her working memory for the steps to swap back the FTS tokens to BNB coins but gave up in fatigue. She decided to get some rest and hoped to see the value back up to $400,000 upon waking; but figure 1 shows the $26.88 current total value of the 32,291,161 FTS from the $1,200 FTS investment transaction. Apparently, there was a smart contract bug in the FTS token that resulted in the balance to display incorrectly that Trust Wallet failed to notify the affected users sooner. If the process of exchanging FTS tokens for BNB coins had not been complicated, she would have been able to save some of the investment by using her working memory to trace back the conversion process as she had done when she swapped the BNB coins to purchase FTS tokens. Instead, she reached a point of surrender from the lack of sleep, clinging to the fleeting hope that the challenging situation would miraculously resolve on its own. Those who were impacted by this incidence suspected whether this was Trust Wallet's intentional dark pattern.

Trust Wallet is a non-custodial cryptocurrency wallet that was designed to be an all-in-one application for the storing or trading of cryptocurrencies, and for users to participate in decentralized finance (DeFi) activities such as staking or interacting with decentralized application (dApps). While the interface of Trust Wallet seems user-friendly, there are various components of this hot wallet that can be overwhelming for all different levels of users: (1) Confusing UI and navigation, e.g., it is not clear of the different purposes of the ‘Swap’ and ‘Browser’ buttons—one allows a swap within the app via DEX and the later enables access of dApps that connect to various decentralized services that also let the users makes a trade. This uncertainty increases the cognitive load and time-on-task leading to fatigue, anxiety, and decrease in motivation that may result in performance degradation. Extensive accommodation would be required even for users with prior knowledge of the crypto systems. (2) With some undefined or unexplained terminologies, novice users may find the jargon in Trust Wallet to be perplexing; this requires attentive decision making, and heuristic-based satisficing may lead to errors. (3) Customer support is only available through the ticket system over email with slow response time can cause fear and increase anxiety. (4) There is no customization available for accessibility, e.g., the option to adjust the interface to help the visually impaired individuals or elderlies to control font size or enlarge features. (5) Confusing security settings to secure users’ wallets properly draw attention to issues of credibility and trust. (6) Although Trust Wallet uses mnemonic backup seed phrase and biometric, it does not have the two-factor authentication (2FA) to ensure credibility and trust further. (7) The app does not have a logout option or button; the user would need a workaround to exit out of the particular wallet by deleting it. Figure 2 shows the screen to log back into the account after the alternative logout solution. The user would need to provide the seed phrase (typically 12, 18, 24 words of private key) that was generated when the account was created, and if lost would also mean losing complete access to the Trust Wallet account; an incidence like this can be detrimental that is cognitively and emotionally taxing for all users.

Conclusion

To this day in May of 2023, Trust Wallet continues to require users to remember elaborate information such as the long login seed phrases and the complicated trading steps that can lead to stress, decision-making mistakes, and site abandonment as they struggle to recall and enter the required details with their abilities. Gaining knowledge on Trust Wallet demands significant accomodation that disruptions can cause loss of attention, and proactive interference can occur where the assumed user interaction does not match with the existing scripts and mental models that are expected for mobile apps. Design team should make efforts to create an intuitive system that provides feedback and prompt supports to align with the human information processing system and to compensate the limitations on capacity, duration, and volatility of their users’ working memory.

References

Adams, E. J., Forsberg, A., & Cowan, N. (2022). The embedded-processes model and language use. In John W. Schwieter & Zhisheng Wen (Eds.), The Cambridge Handbook of Working Memory and Language (1st ed., pp. 73-97). Cambridge University Press. https://www.cambridge.org/core/product/identifier/9781108955638%23CN-bp-5/type/book_part

Atkinson, R. C. & Shiffrin, R. M. (1968). Human memory: A proposed system and its control processes. In Kenneth W. Spence & Janet T. Spence (Eds). Psychology of Learning and Motivation(Vol. 2, pp. 89-195). Academic Press. https://app.nova.edu/toolbox/instructionalproducts/edd8124/fall11/1968-Atkinson_and_Shiffrin.pdf

Baddeley, A. & Hitch, G. (1974). Working memory. In Gordon H. Bower (Ed.) Psychology of Learning and Motivation(Vol. 8, pp. 47-89). Academic Press. https://app.nova.edu/toolbox/instructionalproducts/edd8124/fall11/1974-Baddeley-and-Hitch.pdf

Baddeley, A. (2012). Working memory: Theories, models, and controversies. Annual Review of Psychology, 63(1), 1-29. https://www.annualreviews.org/doi/10.1146/annurev-psych-120710-100422

Barrouillet, P, Bernardin, S., & Camos, V. (2004). Time constraints and resource sharing in adults' working memory spans.Journal of experimental psychology, 133(1), 83-100. https://doi.org/10.1037/0096-3445.133.1.83      

Barrouilleet, P. & Camos, V. (2012). As time goes by: Temporal constraints in working memory. Current Directions in Psychological Science, 21(6), 413-419. http://journals.sagepub.com/doi/10.1177/0963721412459513

Coolidge, F. L., & Wynn, T. (2005). Working memory, its executive functions, and the emergence of modern thinking. Cambridge Archaeological Journal, 15(1), 5–26 https://citeseerx.ist.psu.edu/document?repid=rep1&type=pdf&doi=720bfe4ab13a08a0ced794a4cc20614fee4f0be1

Cowan, N, Morey, C., & Chen, Z. (2007). The legend of the magical number seven. In Sergio Della Sala (Ed.), Tall tales about the mind and brain: Separating fact from fiction (1st ed., pp. 45-59). Oxford University Press. https://doi.org/10.1093/acprof:oso/9780198568773.003.0005

Cowan, N. (2010). The magical mystery four: How is working memory capacity limited, and why? Current Directions in Psychological Science, 19(1), 51-57. http://journals.sagepub.com/doi/10.1177/0963721409359277

Cowan, N. (2014). Working memory underpins cognitive development, learning, and education. Education Psychology Review, 26(2), 197-223. http://link.springer.com/10.1007/s10648-013-9246-y

Cowan, N., Ricker, T. & Vergauwe, E. (2016). Decay theory of immediate memory: From brown (1958) to today (2014). The Quarterly Journal of Experimental Psychology, 69(10), 1969-1995. https://journals.sagepub.com/doi/10.1080/17470218.2014.914546

Csikszentmihalyi, M., Abuhamdeh, S., & Nakamura, J. (2005). Flow. Handbook of competence and motivation, 598-608. https://oleksandr-tereshchuk.com/wp-content/uploads/2021/10/Flow-Book-Excerpts-Oct-13-2021.pdf

Derakshan, N. & Eysenck, M. W. (2009). Anxiety, processing efficiency, and cognitive performance: New developments from attentional control theory. European Psychologist, 14(2), 168-176. https://econtent.hogrefe.com/doi/10.1027/1016-9040.14.2.168

Ericsson, K. A., & Kintsch, W. (1995). Long-term working memory. Psychological Review, 102(2), 211–245. https://doi.org/10.1037/0033-295X.102.2.211

Hattie, J., Hodis, F. A., & Kang, S. H. K (2020). Theories of motivation: Integration and ways forward. Contemporary Educational Psychology, 61(1), 101-865. https://www-sciencedirect-com.ezp.bentley.edu/science/article/pii/S0361476X20300308

Kane, M. & Engle, R. (2002). The role of prefrontal cortex in working-memory capacity, executive attention, and general fluid intelligence: An individual-differences perspective. Psychonomic Bulletin & Review, 9(4), 637-671. http://link.springer.com/10.3758/BF03196323

Manohar, S, & Husain, M. (2016). Working memory for sequences of temporal durations reveals a volatile single-item store. Frontiers of Psychology, 7(1), 1664-1078 http://journal.frontiersin.org/article/10.3389/fpsyg.2016.01655/full

Miller, G.A. (1956). The magical number seven, plus or minus two: Some limits on our capacity for processing information. Psychological Review, 63(2), 81-97. http://psychclassics.yorku.ca/Miller

Nieuwenstein, M. & Wyble, B. (2014). Beyond a mask and against the bottleneck: Retroactive dual-task interference during working memory consolidation of a masked visual target. Journal of Experimental Psychology: General, 143(3), 1409-1427. http://doi.apa.org/getdoi.cfm?doi=10.1037/a0035257

Norman, D. A. (1988). The design of everyday things. The MIT Press. http://kowym.com/wp-content/uploads/2018/08/The-Design-of-Everyday-Things-Don-Norman.pdf

Shipstead, Z. & Engle, R. (2012). Interference within the focus of attention: Working memory tasks reflect more than temporary maintenance. Journal of experimental psychology: Learning, memory, and cognition 39(5), 277-289. https://englelab.gatech.edu/articles/2013/shipsteadengle_2013_interference.pdf

Staal, M. A. (2004). Stress, cognition, and human performance: A literature review and conceptual framework (Publication No. 20060017835). NASA Technical Reports. https://ntrs.nasa.gov/citations/20060017835

Sweller, J. (2003). Evolution of human cognitive architecture. In B. H. Ross (Ed). The Psychology of Learning and Motivation: Advances in Research and Theory(Vol. 43, pp. 215–266). Elsevier Science. https://psycnet.apa.org/record/2004-12443-006

Sweller, J. (2010). Element interactivity and intrinsic, extraneous, and germane cognitive load. Educational Psychology Review, 22(2), 123-138. https://doi.org/10.1007/s10648-010-9128-5

Underwood, B. J., & Postman, L. (1960). Extraexperimental sources of interference in forgetting. Psychological Review, 67(2), 73-95. https://sci-hub.se/10.1037/h0041865