Models of Working Memory

​

The first model proposed by William James (1890) defined the notion of primary memory that holds information in current use. Later in 1968, Atkinson and Shiffrin proposed an alternative model that conceptualized memory in three hypothetical stores, namely, sensory, short term, and long term store (Atkinson & Shiffrin, 1968). A departure from this model was by Craik and Lockhart, 1972, which postulated that memory rather than separate stores, varies along a continuous dimension and depends upon encoding of stimuli (Lockhart & Craik, 1990). In 1974, Baddeley and Hitch postulated the “multi-component model,” replacing the notion of short-term memory.

This model proposed that WM has three major components:

(a) Central executive,

(b) Phonological loop,

(c) Visuospatial sketchpad.

The central executive functions as the control center, directing information between phonological and visuospatial. It focuses and switches attention by activating and integrating information from Long term memory, LTM (Baddeley & Logan, 1999). The phonological loop (PL) holds verbal and acoustic information (Salamé & Baddeley, 1986), stores the sound of language, and prevents it from decay by playing it on rehearsal. The visuospatial (VS) sketchpad stores visual information, creates or revisits mental maps or breaks information into visual subsystems dealing with form, color, texture, and location.

It was further elaborated (Logie, 1995) and divided into two components:

(a) Visual cache, which stores information about form and color and

(b) Inner scribe, which deals with spatial and movement information.

The central executive acts as a supervisory system controlling attention allocation, reasoning, learning, decision making, meta-cognition, and creativity. It directs attention to relevant information, suppresses irreverent information, and actions (Baddeley and Logan, 1999; Ericsson & Kintsch, 1995). In 2000, Baddeley added a fourth component to this model, the “episodic buffer” that provides temporary storage and binds information from its subsidiary systems and from Long Term Memory (Baddeley, 2003). In 1988, Cowan proposed another model and suggested that Working Memory has an attentional system that has the capacity of four chunks (Cowan, 1995). Thus, it can be concluded that working memory can code, manipulate and recombine information at hand to interpret tasks (Miyake & Shah, 1999). As it distributes attention on visual and auditory channels to deliver information efficiently, dual-coded aids creating a number of connections that help to move things from working memory to long-term memory.

​

Limitations of Working Memory

​

Unlike long-term memory, working memory has limitations of capacity and duration for storage, activation and rehearsal. It also faces information decay due to competition for the same space, divided attention in dual-tasks and dissimilarity between visual and auditory signals(Miyake & Shah, 1999). The limitations of working memory are:

​

(1) Limited capacity & duration:  Working Memory is limited in its capacity, denoting that for a task, an individual’s performance decreases with the increase in memory demand. In 1956, Miller provided one of the first quantifications of the limited capacity of working memory. He argued that the limit for ‘one-dimension absolute judgment tasks’, the performance of a person is nearly accurate for somewhere close to seven. Secondly, he argued that the limitation for memory span, i.e a word list that a person can remember incorrect order is approximately equal to seven. He also defined a ‘chunk’ as the largest meaningful unit that a person can recognize (Miller, 1956) and chunking is used to keep groups of information accessible and for easy recall. Miller observed that a memory span can hold a list of seven familiar chunks and stay active for about 20 seconds(Miller, 1956). Experiments conducted by Baddley & Logie on verbal reasoning and comprehension suggested that verbal WM is “ limited in the duration of storage as well as the number of items, a list having no organizing cues can be recalled in 1.5 -2 seconds (Baddeley and Logie, 1999)”. It was noted that after six items there was a degradation in performance. During the process of rehearsal, one part of the sequence could interfere with the memory for the other part and acoustic similarity led to poorer recall of words. Expanding on the concept of chunking, research (Luck & Vogel, 1997) conducted on visual working memory suggested that it can store integrated information for four objects, both about the color and orientation, indicating that visual memory stores integrated objects rather than individual features. Recent research by Cowan, 2001 also implicated an average capacity limit as four chunks. Beyond these limits, items cannot be rehearsed and cannot be retained in long-term memory (Cowan, 2001).

Recognizing the limited capacity of working memory, in 1988, Sweller proposed the cognitive load theory that primarily aimed at reducing the amount of information that can be held in the working memory for learners (Sweller, 1988). He suggested that there are three types of cognitive loads, namely:

(a) Intrinsic load, is inherent and based on the complexity of the topic that cannot be reduced,

(b) Germane load is associated with the construction and processing of schemas

(c) Extraneous load is generated by the manner in which information is delivered to the learner.

Thus, due to the limited capacity of working memory, it is important for designers to reduce the extraneous load as it can affect performance and lead to errors and abandonment (Engle, 2002).

​

(2) Highly Volatile: Another limitation of working memory is that it is highly volatile, when faced with concurrent tasks, an attention conflict is created between the demands of competing tasks leading to error (Kahneman, 1973). Experiments based on word syllables and comprehension indicated that memory span for recall varies with the length of words, this is defined as the word length effect (Baddeley et al, 1975; Cowan, 1992).

However, the memory span is longer for familiar words than completely nonsensical words and can be retained within a duration of two seconds (Baddley, 1999; Logie et al 1996). The experiments also established that “acoustic information is first processed as isolated sounds, then phonemes, then syllables and then turned into sub-lexical and lexical units'' (Logie et al, 2000). As non-words remain to be sub-lexical units, chunking them is not done easily and hence recall to existing semantic networks does not occur. Research also suggests that interference hinders the comprehension of new information leading to two types of errors in serial recall, namely transposition error and intrusion errors. Transposition error occurs when an item from a list is recalled but in wrong order and intrusion error occurs when an item is presented that was not present in the list (Henson, 1998; McCormack et al 2000).

The notion of trace decay (Cowan et al, 1997) suggests that words that are longer in duration and difficult to recall are often faded out from memory unless rehearsed within 18 seconds (Peterson & Peterson, 1959; Ricker et al, 2016). Limitations in visual working memory (Baddeley and Logie, 1999; Logie & Pearson, 1997) include the complexity of information, such as dimensionality, and visual similarity of items to be recalled. It can hold only a single pattern, but the similarity of the patterns can induce limitations on the capacity of the system (Broadbent and Broadbent, 1981; Miyake & Shah, 1999). It is, therefore, important for designers to be mindful of the limitations of working memory while designing products.

​

​

Emotional sub-systems

Emotions are a form of arousal that requires an individual to take action to maintain equilibrium (Duffy,1941; Staal, 2004). Although there are various emotions such as trust, challenge, pleasure, and fear, positive emotions such as motivation direct the executive control towards goal achievement but negative emotions such as anxiety inhibit its functioning.

​

Motivation: Motivation can play a positive role in enhancing the capability of working memory, it improves performance, efficiency and can overcome limitations by guiding, preserving and updating attention towards the goal (Lavie & Fockert, 2005). Research in neuroscience explains the role of dopamine neurons (Daw & Shohamy, 2008) that supports goal-directed behavior led by motivation and reward prediction. Studies suggest that lack of task-related motivation has a negative impact on working memory, whereas (Otsuka Osaka, 2015) moderate levels of anxiety or motivation, which are forms of arousal, can improve performance. But as the as arousal increases it and can also cause a reduction in performance (Yerkes and Dosdon,1908; Staal, 2005).

​

Anxiety: Anxiety in small amounts can help us stay focused but high levels of anxiety can lead to degradation of working memory performance. Studies suggest that the effects of anxiety are proportional to the number of tasks in the central executive( Shackman et al., 2006; Derakshan & Eysenck, 2009).  Individuals undergoing anxiety can be defined as high-trait and high-state (Beck, 1976; Staal, 2004). Individuals with trait anxiety are prone to feel anxious, react faster and perform poorly in every situation whereas individuals with state anxiety only become anxious in some stressful situations (Broadbent & Broadbent, 1988). Experiments also suggest that as an evolutionary advantage, fear may enhance the saliency of an object, where individuals report a ‘pop-out’ pre-attentive detection of the threat stimulus (Carlsson et al., 2004; Triesman, 1988) but it impairs efficiency from the goal (Eysenck & Calvo, 1992). Other factors influencing working memory are age, disabilities, and attentional disorders (Martínez et al, 2020).

​

Design Case: Evaluation of Tele-mental health care company's toolkit

​

Context: XYZ Health is a mental health care service provider offering personalized tele mental health care. The user group considered here is XYZ’s care navigators (CN), they are licensed therapists helping patients navigate the provider network. Their role is to communicate with patients, assess them for severity and keep a tab on their progress. A CN interacts with a patient over a phone call, he/she is required to be able to assess the patient information, take notes about their current mental health status. The patients are undergoing mild to serious issues and at times even have suicidal instincts.

The care navigators are provided with a fragmented tool ecosystem that involves multi-tasking between eight applications simultaneously, for the purpose of this review the comprehensive system is reviewed. It involves:

​

(a) Opening multiple internet browser windows, to access information across platforms (Figure 1)

(b) Using Zendesk, a ticketing tool to access patient’s primary details (Figure 2)

(c) Kareo, a PHR application to store patient data (Figure 3)

(d) Slack, a communication tool for information exchange (Figure 4)

Figure 1: Care Navigator's web browser