Final version: 4
thof July 2014 Supervisor: Frank Nack
The Limits of a Reductionist Approach to Improving
Quality of Sleep through Smartphone Applications
Jochem Ramses Havermans
5965861
Frank Nack,
André Nusselder,
Thesis Master Information Science Human Centered Multimedia
University of Amsterdam Faculty of Science
The Limits of a Reductionist Approach to Improving
Quality of Sleep through Smartphone Applications
ABSTRACT
As persuasive technology relies on a reductionist approach that borrows from positivistic psychology, the wicked problems it addresses are reduced to quantifiable models. Although wicked problems are by definition not finitely definable, the solutionist approach of persuasive technology rules out alternative views on definitions and solutions to a wicked problem. This paper exemplifies this scientific problem through a paradigmatic case study of two smartphone applications that aim at measuring and classifying sleep quality in order to improve it. A narrative analysis of four academic papers, three interviews with experts on the field of sleep disorders, and six interviews with users of the applications (of which two suffer from a diagnosed sleep disorder), shows that poor sleep can indeed be considered a wicked problem, and that even though these applications claim to address a wider issue, they in fact only focus on a small aspect of poor sleep. The narrative analysis also shows that there are different approaches to solving poor sleep that address a wider spectrum of the wicked problem than the approach of the two applications: raising awareness on sleep hygiene and restricting sleep time seem to be thought of as more successful methods in the solution to poor sleep.
Keywords
Persuasive technology, sleep quality, solutionism, reductionism, constructivism, wicked problems, design thinking, narrative analysis.
1. INTRODUCTION
Descending from Descartes’ introduction of the mechanistic worldview, there has been a strong tradition in the field of psychology that views the human body as a machine [2]. From this point of view, the way a human body works is explained along the same lines as one would explain the mechanism of a machine, such as a clock. Although, in his mechanistic worldview, Descartes’ initially made a distinction between the immaterial (spiritual) mind and the material (mechanical) body [6], a shift towards interest in computational theory in the early 1950s has led to the mechanical modeling of the cognitive processes of the mind. This development allowed for a new reductionist view in which (also) the minds of humans are explained along the same lines as their mechanical (virtual) counterparts [2].
As this positivistic reductionist approach to studying and explaining human behavior through cognitive processes is still widespread in psychology, persuasive technology has adopted this approach in its urge to fix (social) problems [10], such as obesity, food waste, or criminal endeavor. Morozov describes how this ideology of persuasive technology –– which he calls “solutionism” –– aims to improve the world by “[r]ecasting all complex social situations in either as neatly defined problems with definite, computable solutions or as transparent and self-evident processes that can easily be optimized […]” [10, p. 5]. The problem with technology’s solutionism is that it tends to close off the definition of a (social) problem. By building a computationally quantified model of the problem and its solution,
alternative models and solutions are ruled out. Whereas in fact the social problem addressed, is likely to be complex to such an extent that it is by definition not finitely definable: new ways of viewing and explaining a problem emerge as time continues. This type of social problems are referred to as “wicked problems” [11]. For the sake of solvability, the solutionist approach in the development of persuasive technology tends to reduce a wicked problem to merely its measurable quantitative aspects and therefore potentially rules out conflicting perspectives on the wicked problem and/or ignores the contextual qualitative aspects of the problem [10].
A similar problem is detected in positivist psychology’s reductionist approach to cognitive processes, where social problems are studied and explained in isolation, instead of being studies and explained through a context-sensitive holistic approach [2]. As with the solutionist approach to persuasive technology, the reductionist approach to cognitive processes, leads to simplified models of human behavior. These models rule out conflicting views on the problem they are dealing with. This paper will exemplify the scientific problem of solutionism through a paradigmatic case study of two applications that claim to improve quality of sleep: Sleep Time1 (see Image 1) and Sleep
As Android2 (see Image 2).
In order to determine to what extent these applications are addressing the (wicked) problem they claim to address (improving quality of sleep), this paper will first explore the different ways in which the problem can be defined through a narrative analysis of academic papers, interviews with experts on the subject of poor sleep and its solutions, patients suffering from sleep problems, and users of both applications. The narrative analysis focuses on what stories are told by the different narrative accounts to describe and understand the wickedness of poor sleep [12]. The choice for this type of analysis is inspired by Bruner’s call for a narrative approach –– his reaction to the (reductionist) logical-scientific approach in psychology [3]. The narrative approach lets go of the logical-scientific objective to find and know the truth (of a social problem), and replaces it with an approach that embraces different versions of the truth (in order to match the complexity of a social problem) [16]. The narrative analysis of this paper will first serve as a means to explore the wicked problem (section 3) and offer a (indefinite) definition of the problem space through convergence. The second half of the narrative analysis explores possible solutions to the issue of poor sleep and points out how the two applications (Sleep As Android and Sleep Time) only address a small aspect of the wicked problem (section 4). Both applications, namely, aim at measuring and classifying quality of sleep in order to improve it. By measuring and classifying in which state of sleep the user is, the applications pick the best moment in the time frame of half an hour (indicated by the user) in which to wake the user. The applications, thus, essentially function as a so-called Bio Alarm Clock. This quantitative reductionist approach of both
1 http://www.azumio.com/apps/sleep-time/
applications to solving a wicked problem is the reason for which these applications are suitable examples of the overarching scientific problem. As this paper shows, these two applications claim to solve a wicked problem, while indeed only addressing a simple sub-problem.
Image 1 Sleep Time Screenshot Image 2 Sleep As Android Screenshot
This paradigmatic case study of Sleep As Android and Sleep Time only serves to exemplify and illustrate the overarching scientific problem of persuasive technology’s reductionist approach to the solution of wicked problems. Although this study helps to give insight in a common problem, it would be wrong to generalize from its findings. The choice for this particular research set-up agrees with the opinion that there is not one set of rules by which all wicked problems should be explored and solved [11]. As also pointed out by Morozov, each possible solution to each social problem calls for its own approach and the way in which technology can help with a possible solution should be considered case-specifically [10].
2. RELATED WORK
Although the narrative approach might not yet be as widely adopted in psychology as more reductionist approaches, Bruner’s narrative approach has already began to find its way to IT development in the disguise of “Design Thinking” [9]. This concept encourages a thorough exploration of both the ways in which a problem can be defined (the problem space) and the ways in which a problem can be addressed (the solution space), through an iterative approach [15]. Design Thinking thus offers an alternative to more rigid IT development concepts in which problems are strictly defined at the start of a development process. Most importantly, the Design Thinking approach to exploring the problem and solution space is suitable for dealing with ambiguous wicked problems.
The notion of wicked problems stems from the area of policy sciences, in which Rittel & Webber claim that strict scientific approaches to solving (wicked) problems (of social policy) are by definition inadequate, as it is impossible to fully define such a problem [11]. They state that we live in a world where each problem can be viewed from any number of perspectives, of which none are inherently false or correct. This also implies that “there are no solutions in the sense of definitive and objective answers” [11, p. 155]. Rittel & Webber have condensed the complex nature of these wicked problems into ten characteristics:
“1. There is no definitive formulation of a wicked problem; 2. Wicked problems have no stopping rule;
3. Solutions to wicked problems are not true-or-false, but good-or-bad;
4. There is no immediate and no ultimate test of a solution to a wicked problem;
5. Every solution to a wicked problem is a "one-shot operation"; because there is no opportunity to learn by trial-and-error, every attempt counts significantly;
6. Wicked problems do not have an enumerable (or an exhaustively describable) set of potential solutions, nor is there a well-described set of permissible operations that may be incorporated into the plan;
7. Every wicked problem is essentially unique;
8. Every wicked problem can be considered to be a symptom of another problem;
9. The existence of a discrepancy representing a wicked problem can be explained in numerous ways. The choice of explanation determines the nature of the problem's resolution;
10. The planner has no right to be wrong.” [11]
These statements made by Rittel & Webber link both to the narrative approach of Bruner, in which he claims that there is not just one reality, as it is constructed through different (potentially conflicting) narratives [4], and Morozov’s criticism of solutionism in technology, which critiques the current urge to technologically solve problems that don’t exist; things are being “solved”, but they are only remotely related to the actual wicked problem a technology aims to solve. Bruner’s ideas are built on his earlier contributions to constructivist theory, where he elaborated on the idea that new learning experiences are constructed by the learner’s past [5]. This stance fits into the idea that there is no such thing as one objective definition of a wicked problem, as the background of the person defining the wicked problem will always add a layer of subjectivity.
The fact that Design Thinking is implemented in IT-contexts [9], also implies that not every new technology that addresses a wicked problem is of the solutionist type. In fact, there are smartphone applications addressing the very issue of poor sleep in a different way than Sleep As Android and Sleep Time. An example is the Dutch smartphone application Slaap Lekker3. This
application tries to gradually improve the user’s quality of sleep by giving the user a set of (reflective) assignments each week. These assignments vary from frequently ventilating one’s bedroom to reflecting on a manually filled out sleep journal. What makes this application different from Sleep As Android and Sleep Time is that it focuses on changing the user’s behavior by letting the user reflect of him- or herself, instead of algorithmically measuring and analyzing the user’s behavior through sensors. As Design Thinking takes the notion of wicked problems central in its methodology, and translates the narrative approach into an IT-context, I have chosen to adopt its distinction of the exploration of problem space and the exploration of solution space in this study’s narrative analysis. In this way, this paper will
show how a narrative approach in the form of Design Thinking can qualitatively give insight into a wicked problem and its solutions.
While this paper’s narrative analysis will show that poor sleep is indeed a wicked problem, it is important to note there exist several rigid definitions of certain types of poor sleep. The two most commonly used definitions among the Experts interviewed in the narrative analysis of this paper, are the DSM-IV-TR definition and ICSD (International Classification of Sleep Disorders). The DSM-IV-TR definition of Insomnia4 is summarized by Roth (in
Paper D of the narrative analysis) as “(1) difficulty falling asleep, staying asleep or nonrestorative sleep; (2) this difficulty is present despite adequate opportunity and circumstance to sleep; (3) this impairment in sleep is associated with daytime impairment or distress; and (4) this sleep difficulty occurs at least 3 times per week and has been a problem for at least 1 month” [13, p. 1]. The, more intricate, International Classification of Sleep Disorders (ICSD) “was produced primarily for diagnostic and epidemiologic purposes so that disorders could be indexed and morbidity and mortality information could be recorded and retrieved.” [1, p. 18] Its intricacy makes it difficult to summarize; for the purpose of this paper it is sufficient to note that the ICSD is mainly an important resource for academic research on sleep disorders [1]. The Papers used in this narrative analysis each deal with different aspects of the issue of poor sleep. Paper A (“The Subjective Meaning of Sleep Quality”) exemplifies how “sleep quality” is an ill-defined term; although it is used widely to determine insomnia, it is used arbitrarily by researchers, clinicians and patients. The paper’s aim is “to investigate the subjective meaning of sleep quality among individuals with insomnia and normal sleepers.” [7, p. 383]. Their conclusion is that sentiment after waking up and sentiment throughout the day are perceived as important subjective measurements for determining sleep quality [7]. Paper B (“Self-help therapy for insomnia: a meta-analysis”) finds that self-help-treatments to insomnia only offer small to moderate beneficial effects, but can serve as a useful addition to other forms of treatment [14]. Paper C (“Recent developments in home sleep monitoring devices”) discusses several new forms of home sleep monitoring and points out that while their popularity continues to rise, academic validation of these devices is still in an infant stage [8]. Paper D (“Insomnia Definition, Prevalence, Etiology and Consequences”) is useful in this narrative analysis for its earlier quoted definition of insomnia.
3. EXPLORATION OF THE PROBLEM
SPACE
Before diving into the solution space of the issue of poor sleep, this section will begin to show to what extent the issue of poor sleep is indeed a wicked problem (as according to Rittel & Webber’s ten characteristics of wicked problems [11]). In this section, a collection of four academic papers [7, 8, 13, 14], three semi-structured interviews with experts on the field of sleeping disorders (a psychologist, a neurologist/somnologist, and a chronobiologist), four semi-structured interviews with (ex)-users of Sleep As Android (of which two participants had a diagnosed sleeping disorder), and two semi-structured interviews with users of Sleep Time give insight into what poor sleep is and what causes
4..http://www.hcp.med.harvard.edu/wmh/ftpdir/affiliatedstudies_B
IQ_algorithm.pdf
it. Dedoose5 has been used for the coding and analysis of the total
of thirteen narrative accounts.
In order to make sense of the collection of narrative accounts, a sequential analysis was performed, in which the different narrative accounts were broken into excerpts, which were assigned a code belonging to a corresponding category [12]. These categories represent a narrative in which the excerpts fit; similar narratives by different narrative accounts are assigned to the same category. For example: the definitions of insomnia mentioned in Paper A and Paper D both fit into the DSM-definition and are therefore assigned to the same category (see Table 1). By determining which types of narratives are told by which narrative accounts, the wicked problem is explored and frequently reoccurring stories surface. In this way, this paper shows which types of definitions and solutions seem to be preferred by the different narrative accounts. The Tables in this paper present either whether a certain narrative was present in a narrative account (for example in Table 1), or how often a certain narrative was mentioned in a narrative account (for example in Table 2). The initial categories are based on academic literature on the subject of poor sleep, and crosschecked by one of the interviewed experts. Subsequently the categories were updated and fine-tuned after the collection and processing of each new account presented alternative narratives.
3.1 What is Poor Sleep?
In the narrative accounts, three types of definitions of poor sleep came forward. Table 1 shows, per narrative account, which types of definition of poor sleep were used/mentioned. Paper B and Paper C are not present in Table 1 as they do not discuss definitions of poor sleep. Firstly, the DSM definition for Sleep-Wake disorders was referred to by Paper A, Paper D, Expert A, and Expert B (as shown in Table 1). Expert A and Expert B pointed out that they used the DSM definition in combination with the ICSD (International Classification of Sleep Disorders) for research purposes. Expert C, the only medical practitioner of the three experts, on the other hand does not make use of these strict definitions and relies on more subjective means to determine whether or not a patient has a sleeping disorder or not. His motivation for this is that, when a patient says he or she has problems sleeping, this perceived problem needs to be solved; whether this problem is also a problem according to the definitions of DSM and ICSD is less relevant.
Not surprisingly, none of the interviewed Users of Sleep Time and Sleep As Android define poor sleep or their sleep disorder along the strict thresholds of definitions such as DSM and ICSD. Instead of using rigid definitions, poor sleep to them is defined on the basis of subjective perception of their sleep quality (User B: “[…] when I dream intensely about the things that occupy my mind and when they eventually merge with the space in which I sleep.”), their sleep quantity (User E: “[…] when you wake up a lot of times during the night […]”), or their performance the day thereafter (User C: “Poor sleep is when I can no longer concentrate around dinner time.”). These subjective definitions vary greatly in their content.
A lack of agreement on what poor sleep is, is backed by Paper A’s quotes of Akerstedt, Hume, Minors, and Waterhouse (“there seems to be very little systematic knowledge as to what actually
5 http://www.dedoose.com/ –– a web-based application for the
constitutes subjectively good sleep and how this should be measured”) and Buysse et al. who referred to sleep quality as a “complex phenomenon that is difficult to define and measure objectively.” [7, p. 383].
Table 1 Definitions of poor sleep per narrative account
Definitions of Poor Sleep DSM ICSD Subjective Definition
The Subjective Meaning of Sleep Quality: A Comparison of Individuals with and without Insomnia (Paper A)
1 1
Insomnia Definition, Prevalence, Etiology, and
Consequences (Paper D) 1 Psychologist (Expert A) 1 1 Chronobiologist (Expert B) 1 1 Neurologist (Expert C) 1 Ex-User Sleep As Android (User A) 1 Frequent User Sleep As Android (User B) 1 Frequent User SleepTime (User C) 1 Novel User Sleep As Android (User D) 1 Novel User Sleep As Android (User E) 1 Novel User Sleep Time (User F) 1 Totals 4 2 8
Table 2 Mentions of types of poor sleep per narrative account
Types of Poor Sleep Circudian Rythm Sleep Disorder (CRSD) General Poor Sleep Insom -nia Sleep Apnea
The Subjective Meaning of Sleep Quality: A Comparison of Individuals with and without Insomnia (Paper A)
4
Self-help therapy for insomnia: A meta-analysis (Paper B)
2 Recent developments in home sleep
monitoring devices (Paper C) 1 3 3 Insomnia Definition, Prevalence,
Etiology, and Consequences.doc (Paper D)
2
Psychologist (Expert A) 3 9 1 Chronobiologist (Expert B) 6
Neurologist (Expert C) 8 6 4 Ex-User Sleep As Android (User A) 1
Frequent User Sleep As Android (User
B) 3 6
Frequent User SleepTime (User C) 3
Novel User Sleep As Android (User D) 2 Novel User Sleep As Android (User E) 2
Novel User Sleep Time (User F) 3
Totals 21 9 34 8
Another thing that came forward from the narrative analysis is that, in the narrative accounts, there are distinctions between different types of poor sleep. In the narrative accounts, a distinction can be made between four different types of poor sleep (see Table 2), of which three are medically classified as sleep disorders (CRSD, Insomnia, and Sleep Apnea). Table 2 shows how often a type of poor sleep was mentioned by a narrative
account. For example: User B mentioned insomnia in six narrative excerpts and CRSD in two narrative excerpts. This tells us User B suffers from both insomnia and CRSD, but seems to be more focused on his or her insomnia.
Whereas the Papers and Experts solely focus on the three types of medical sleep disorders, the issues of the majority of Users with poor sleep can be classified as general poor sleep. Only User B and User D claim to have a medically diagnosed sleep disorder. This distinction is important to take into account as different types of poor sleep generally have different causes (as explained in section 3.2) and different solutions (as discussed in section 4).
3.2 What Causes Poor Sleep?
In order to understand what the different types of poor sleep entail––and to eventually prevent or solve them––, it is essential to look at their causes. As mentioned, the narrative accounts made a distinction between four types of sleep disorders, of which the Circudian Rhythm Sleep Disorder (CRSD) can be divided into a primary disorder (referring to a congenital type of CRSD) and a secondary disorder (referring to a circumstantial type of CRSD). A code-occurrence analysis of the total of thirteen narrative accounts shows the difference in the nature of the four types (and two sub-types) of poor sleep (see Table 3). Table 3 shows how often a mentioned cause of poor sleep was mentioned in combination with a certain type of poor sleep in one narrative excerpt.
Table 3 Code co-occurrence of poor sleep causes and poor sleep types CRSD Pri mary CRSD Secon dary CRSD General Poor Sleep Insomnia Sleep Apnea Biological Causes 8 3 5 3 5 1 Congenitally Late/Early Biological Clock 3 3 Disturbed Biological Clock 5 5 1 Hyperactivity 1 Lack of Melatonin 1 Multiple Sclerosis 1 Quitting Smoking 1 Alcohol Abuse 1 2 Respiratory Disorder 1 Environmental Causes 3 Bed Partner 1 Heat 1 Noise 1 Psychoscocial Causes 5 5 6 25 Anxiety 1 1 8 Depression 4
Poor Sleep Hygiene 3 3 2 5 Stress 1 1 4 8
Primary CRSD, for example, was solely mentioned three times in combination with a congenitally late/early biological clock. From this can be concluded that Primary CRSD is mainly caused by a congenitally late/early biological clock. One example of such a narrative excerpt in which Primary CRSD was mentioned in combination with a congenitally late/early biological clock is
Expert B, who states that “Primary CRSD is caused by an internal defect of the biological clock”, whereas Secondary CRSDs “ […] are more the result of an externally imposed change to one’s rhythm, for example by night shifts, a jet lag, blindness, or dementia. In those cases there is also a problem with the patient’s rhythm, but its cause is external” instead of internal.
As Table 3 shows, general poor sleep, the least strictly definable type of poor sleep, can be the result of any number of causes. Causes of poor sleep for those Users suffering from general poor sleep vary from “[…] a neighbor who drives off on his motorcycle every morning at 5AM […]” (User A) to “I quit smoking two years ago” (User C). Most frequently mentioned causes, however, are psychosocial causes, such as stress (cognitive arousal): “In general [I sleep poorly] when I have had a really busy week. So when I haven’t had a lot of worries, I tend to sleep better.” (User C).
These psychosocial causes are also frequently mentioned in combination with Insomnia. Insomnia––essentially being an advanced version of general poor sleep [13]––is associated with more grave psychosocial causes, such as anxiety and depression. Insomnia can also be the result of a number of biological causes, among which hyperactivity, the effects of alcohol abuse, and, in the case of User D, multiple sclerosis.
The cause of Sleep Apnea, the least mentioned type of poor sleep, was only explicitly mentioned in combination with a cause once, but seems to be linked to a respiratory disorder (Expert C). Something that the analysis of the causes of poor sleep also brought forward is that a number of causes of poor sleep can also act up as symptoms of poor sleep. User B gave one such example: “[An effect of sleeping poorly] is that the day thereafter I was usually afraid to go to sleep. I have had periods in which I did such strange things during my sleep I did not dare to go to sleep: either because of the dreams or the things I would do during them. So, that does not really contribute [to the solution of poor sleep]”. In User B’s case, anxiety was both a cause and symptom of poor sleep.
3.3 Problem Space Convergence
Although clear definitions of types of poor sleep do exist (such as the DSM definition, and the ICSD definition), they seem to be limited to actual sleep disorders and are mainly used for research purposes. From the viewpoint of both a practitioner and all users, subjective accounts of quality of sleep seem to be leading in defining what is poor sleep and what is not. As the definition of what is perceived as poor sleep differs greatly among the narrative accounts, no general definition of what poor sleep is can be formulated. From this can be concluded that there is no final definition of the problem “poor sleep”. In fact, poor sleep seems to be a collection of wicked problems. This is in correspondence with Rittel & Webber’s first (of ten) characteristic of wicked problems [11, p. 161]. Of course, most accounts of what poor sleep bear some similarity and some trends can be found, but the accounts never fully overlap. As best put by Rittel & Webber: “7. Every wicked problem is essentially unique” [11, p. 164]
The wide variety of mentioned causes for sleeping poorly knows no sharp distinction with the symptoms of poor sleep. The vicious loop exemplified by the account of User B, links to the eighth characteristic of wicked problems [11, p. 165]. At the same time, each different cause offers a new explanation of what poor sleep is. Which cause is thought to be valid also has consequences for which solution is chosen (as section 4 will show). However, there is no formula to find out which cause or solution to poor sleep is
the “right” one. This phenomenon is explained by the ninth characteristic of wicked problems [11, p. 166].
4. EXPLORATION OF THE SOLUTION
SPACE
Moving to the solution space of the narrative analysis offers an answer to the questions to what extent Sleep Time and Sleep As Android measure and address the problem of poor sleep. In order to do this, the documentation of both applications is incorporated in the narrative analysis from here onwards, and will be referenced to in the footnotes.
This part of the narrative analysis first extracts all mentioned measurements for determining quality of sleep and all mentioned solutions to the issue of poor sleep by the different narrative accounts (section 4.1.1 and section 4.2.1). Comparing the measurements and solutions surfaced by this part of the narrative analysis to the measurements and solutions used by Sleep Time and Sleep As Android sheds light on the discrepancy between the complexity of the wicked problem poor sleep poses and the simplicity of the offered the measurements and solutions by the two applications (section 4.1.2 and section 4.2.2).
4.1 How Can Poor Sleep Be Measured?
4.1.1 All Measurements Mentioned in Papers,
Experts, and Users
In order to solve a problem, one first needs to determine there is a problem by measuring it in some way. The analysis of the narrative accounts shows a large number of methods, divided into objective measurements and subjective measurements (as shown in Table 4). Table 4 shows how many narrative excerpts per user group mention using a certain type of measurement to determine quality of sleep. For example, using measurements made by home sleep-monitoring devices based on movement was mentioned 11 times by the Users, twice by the Experts and once by the Papers. Although the list in Table 4 is surely not exhaustive, it does give insight into the possible methods to measure quality of sleep. Due to the nature of this research, all Users use a Home Sleep-Monitoring Device based on movement (namely, Sleep Time or Sleep As Android); hence it is mentioned often as a used measurement for determining quality of sleep. However, none of the users solely base their sleep performance on the results of the measurements of these apps. As also shown in Table 1, all users base sleep performance on subjective factors.
Interestingly, the Experts share with the Users that they combine objective measurements and subjective measurements (as shown in Table 5). Table 5 shows how often the Papers and the Experts mention using an objective or a subjective form of measurement of sleep quality. In all narrative accounts (except for Paper B in which measurements are not discussed) objective measurements and subjective measurements are used in combination as a form of validation. Expert C, for example, mentions that a patient’s story is central in measuring sleep performance: “and if a patient tells he or she has a sleep problem, then that can be objectified using objective measurements”. Expert A also measures sleep quality subjectively: “So, what people say about how they feel during waking up and how they experienced the night, but besides that I also make use of actigraphy in my research”. An important reason for combining subjective and objective measurement is best described by Expert B “there are people who complain vigorously about their sleep and if you measure their sleep with EEG it turns out that objectively their sleep quality is near perfect… but they
say they haven’t slept a minute. Well, then there is still something wrong. What is wrong, we might not know exactly, but the subjective judgment on whether you sleep well or not, is very crucial.”
Table 4 Used Measurements for Quality of Sleep by Narrative Accounts
Papers Experts Users
Objective Measurements 8 10 11
o Actigraphy 3
o Home Sleep-Monitoring Devices 4 2 11
§ Based on Autonomic Signals 1
§ Based on Brain Activity Signals 1
§ Based on Movement 1 2 11
§ Correlated with Longitudinal Data 1
o Measurement of Melatonine Production 1
o Polysomnography (PSG) 3 4
§ EEG 1 4
§ EMG 1
§ EOG 1
Subjective Measurements 6 7 13
o Perceived Performance in Daily Life 2 4
o Perceived Sleep Quality 2 4 3
o Perceived Sleep-Wake Rhythm 3 3 1
o Sentiment After Waking Up 1 2 3
o Sentiment When Trying To Fall Asleep 1 2
Table 5 Mentions of Objective/Subjective Measurement Methods per Narrative Account
Definitions of Poor Sleep Objective
Measurements Subjective Measurements
The Subjective Meaning of Sleep Quality: A Comparison of Individuals with and without
Insomnia (Paper A) 2 4 Recent developments in home sleep monitoring
devices (Paper C) 5 1 Insomnia Definition, Prevalence, Etiology, and
Consequences.doc (Paper D) 1 1 Psychologist (Expert A) 2 3 Chronobiologist (Expert B) 3 1 Neurologist (Expert C) 5 3 Totals 18 13
4.1.2 Measurements Used by Sleep Time & Sleep As
Android
Sleep Time and Sleep As Android both make use of an actigraphic technique for determining, sleep-wake rhythm, depth of sleep, and thereby quality and efficiency of sleep. This actigraphic measurement is based on movement detected by the accelerometer in the smartphone of the user. For accurate measurement the smartphone is to be placed face down on the user’s mattress
during the night6. This type of measurements fits into subcategory
“Based on Movement” in the category “Home Sleep-Monitoring Devices” under “Objective Measurements” (as in Table 4). Whereas Sleep Time only uses this objective measurement method, Sleep As Android offers the user to rate their “Perceived Sleep Quality” (as in Table 4) on a five-star scale. This feature serves as a means for users to find correlations between the objective measurements and their subjective perception of sleep quality 7.
The strong focus of both applications on objective measurements shows a discrepancy with the results of the narrative analysis. As far as it comes to measuring sleep quality it is safe to say that there seems to be agreement among Papers, Experts and Users that objective measurements alone are not enough to determine whether someone is sleeping poorly. Subjective perception plays an important part on the side of the person suffering from poor sleep. A combination between subjective and objective measurements seems to be the most agreeable option among the narrative accounts.
Sleep Time, however, only measures sleep quality objectively and fails to include the user’s perception of sleep quality in its measurements. Sleep As Android, on the other hand, does offer the user an option of correlating subjective perception with objective measurement. This option was however only mentioned once by User B, and in a negative sense. User B mentioned using the rating feature for assigning a subjective judgment to a night’s sleep. This User’s rating, however, did not seem to correlate with the objective measurements: “first you rate your night and afterwards you are presented with the objective measurements; and then it turns out the subjective ratings are completely random”. The user partly assigns this discrepancy to an irregular sleeping pattern as a nurse.
All Users and Experts interviewed for this paper were asked to give their opinion on the measurements performed by the sleep application they were using. Due to the modest size of this qualitative study the results of the answers to these questions should not be interpreted as a means to determine which of the two applications is deemed to perform better. The results, however, do give an insight into reoccurring views and objections to the measurements used by both applications. Table 6 gives an impression of the most heard narratives concerning the objective measurement based on movement, found in the nine interviews. Two of the Experts and three of the Users share a general distrust towards the accuracy of measurement based on movement by an accelerometer. Whereas Expert B has plans to validate this type of Home-Sleep Monitoring Device against professional PSG-measurements, Expert C does not believe it is “[…] a very reliable
instrument” based on experiences with patients.
Users D, E, and F, base their distrust towards the accuracy of their application’s objective measurements on a discrepancy between their subjective collection of their sleep-wake rhythm and the graphed results; i.e. “given that I knew that I was awake several times and I saw that [the application] didn’t detect it, I don’t know what to say about the other patterns” (User E).
Having a smartphone in bed proved distracting from sleep in the cases of User B, E, and F for different reasons. User B says:
6 https://sites.google.com/site/sleepasandroid/doc/background +
http://www.azumio.com/apps/sleep-time/
“there are moments at which I am bothered [by my phone] because it heats up when it is fully charged. So when I am already uncomfortably warm in bed, I put it away. That means that sometimes it will wake me up [during the night]”. User E said to have woken up wondering whether the phone had crept under the sheets. User F claimed to think about what the application was measuring with every movement, which was distracting the user from falling asleep.
Expert C, and User C and D report a lack of context-awareness on the side of the monitoring device as a deficit of the applications. This objection comes down to the argument that the application might be able to say that a user has slept poorly, but does not have the data to show why that is. As User C puts it: “So yeah, it helps because you can see that you slept like crap that night. That could be because my flat mate came home drunk, but there is no way to know that [based on the applications measurements]”.
Reports on influence of a phone’s radio signals on brain activity are a reason for User A and F to no longer use the application: “isn’t placing a phone next to your head always said to be bad for you? […] The phone’s antenna receives wave signals […], which stimulates certain brain activity. I think it’s a bit contradictory to then place a phone in bed with the idea to sleep better” (User F). User A and F seem to also share the sentiment that sharing a bed with someone leads to inaccurate results. Although Sleep Time’s documentation states that they “developed a very sensitive algorithm for sleep time that, if placed close enough to you, will be isolated enough from their movement that it won’t skew the measurement”8 and Sleep As Android’s website claims that “it
could still work assuming you keep your phone close to your body and ideally on your side of the bed”9, these two Users are not
convinced, based on their own experiences.
Table 6 Reoccurring views on objective measurement based on movement
What these opinions and experiences of the Users and Experts show is that even if the concept of waking up during light sleep is more optimal than waking up during deep sleep (which research has yet to prove [8, p. 8]), it is still questionable whether the objective measurements performed by a smartphone’s accelerometer are capable enough of making a distinction between different sleep phases. Besides that, having the phone in bed seems to also have negative effects towards sleep quality as it proves to be a distracting factor for some Users, while others are afraid of possible negative effects of being exposed to the phone’s
8 http://www.azumio.com/apps/sleep-time/
9 https://sites.google.com/site/sleepasandroid/q-a-faq
radio signals. The lack of contextual (longitudinal) data is also viewed as a reason to doubt the effectiveness of the use of objective measurements alone. To state the user is in light or deep sleep alone does not say much: in order to address the problem the user also needs to understand why he or she is sleeping unsatisfactory. In other words: in order to understand the meaning of the data based on the objective measurements, they need to be placed in a context.
4.2 How Can Poor Sleep Be Prevented or
Solved?
While the different measurement techniques in the last section potentially offer insight into the quality of a person’s sleep, this section discusses the steps that can be taken to address the causes of poor sleep by preventing or solving it.
4.2.1 All Solutions Mentioned in Papers, Experts,
and Users
Table 7 gives an overview of all the solutions to poor sleep mentioned by the narrative accounts and offers an insight in how often these solutions to poor sleep are mentioned in combination with a type of poor sleep. The mentions counted in this part of the narrative analysis are only those mentions in which the narrative account was positive towards that type of solution. For example: in all narrative accounts restriction of sleep time was mentioned 6 times in total as an effective solution to Insomnia.
Table 7 Positive mentions of solutions in combination with a type of poor sleep
CRSD General Poor Sleep Insom-nia Sleep Apnea
Bio Alarm Clock 1 1
Black-out Curtains 1 1
CPAP 1
Cognitive Behavioral Therapy 8 8 27
Behavioral Techniques 5 8 17
Raising Awareness on Sleep Hygiene 2 6 6 Relaxation Exercises 1 3 Restriction of Sleep Time 3 1 6 Stimulus Control 2 Cognitive Techniques 3 10 Cognitive Restructuring 2 4 Imagery Training 2 Paradoxical Intention 2 Sleep Diary 1 2 Light Therapy 3 Medication 2 4 Benzodiazepines 1 4 Melatonin 1 Music Therapy 2 Physical Exercise 1
The subclass “Behavioral Techniques” of class “Cognitive Behavioral Therapy” was mentioned often for three varieties of poor sleep (being CRSD, General Poor Sleep, and Insomnia). The most mentioned successful solutions to poor sleep are “Restriction of Sleep Time” and “Raising Awareness on Sleep Hygiene”. Restriction of sleep time is focused on “cutting down the time
Experts Users
A B C A B C D E F
General distrust towards accuracy of measurement
1 1 1 1 1
Having a phone in bed can be distracting
1 1 1
Lack of contextual/longitudinal data 1 1 1
Radiation from phone might interfere
with sleep quality 1 1
Sharing a bed leads to inaccurate
spent in bed to the time that someone actually sleeps” (Expert A). Raising awareness on sleep hygiene is about becoming aware of the little things that affect sleep quality, such as “only using your bed to sleep in, slowing down at the end of the day, and avoiding naps during daytime. Those are very useful rules of sleep hygiene that in theory everyone should adhere to” (Expert B). These two solutions seem to reach across different types of poor sleep and, because of this broad reach, could potentially match the complexity of the issue of poor sleep.
The other solutions mentioned as being useful by the narrative accounts, seem to address only one or two of the types of poor sleep. Therefore, it can be said that the other solutions mentioned do not offer broad solutions to the general issue of poor sleep. CPAP, for example, is only mentioned as a possible solution to Sleep Apnea, and a Bio Alarm Clock was mentioned once as a possible solution to General Poor Sleep and once as a solution to Insomnia.
4.2.2 Solutions Offered by Sleep Time & Sleep As
Android
Sleep Time and Sleep As Android pose a number of solutions to the issue of poor sleep. Some solutions are explicitly mentioned on their website (such as the Bio Alarm Clock) and other solutions are implicit in the sense that the Users interviewed for this paper believed that the applications helped them in a way that reached beyond the explicit features of the applications (such as raising awareness on sleep hygiene). All Users interviewed for this paper were asked about their views on the solutions offered by Sleep Time and/or Sleep As Android. Table 8 shows whether an either explicit or implicit solution was mentioned as effective by a narrative account.
The core solution that Sleep Time and Sleep As Android are built on is the use of a Bio Alarm Clock. Sleep As Android is said to “[use] the sleep phase information based on accelerometric sensors in your mobile phone to recognize your sleep phases. In general you specify a time widow during which you wish to be woken up and the Smart wake up algorithm looks for significant light sleep indications to trigger the alarm. Waking up in a light sleep phase on the other hand feels natural and is similar to the experience of waking up on the weekend without an alarm clock at all.”10
Throughout all the narrative accounts, the Bio Alarm Clock is however only mentioned twice as a useful solution to either insomnia or general poor sleep. User B found that the Bio Alarm Clock really helped him to wake up, whereas User A said that, although it did not work for him personally, he believes the Bio Alarm Clock might work for other people.
The lack of enthusiasm for the Bio Alarm Clock is somewhat backed by Paper C’s comments on these types of Smart Alarms: “There is clearly a sense of “face validity” for the concept of an alarm clock allowing one to wake up at the optimal time, that is, when sleep is already lightest. […] The lack of data is concerning, given that this feature is no doubt an important attraction to potential consumers. In fact, how the stage of sleep from which one awakens impacts subjective alertness remains largely unknown.” [8, p. 8].
On top of that, there is an obvious discrepancy between the complexity of the issue of poor sleep and the simplistic approach of the Bio Alarm Clock. The Bio Alarm Clock only focuses on the
10 https://sites.google.com/site/sleepasandroid/doc/background
end of a user’s sleeping pattern (right before waking up) in order to determine the moment at which the user sleeps the lightest and thus should be woken up. This is a good example of how both applications try to solve the wicked problem of poor sleep, by only focusing on a sub-problem: the user’s sentiment after waking up. As we have seen in section 4.1.1, sentiment after waking up can be indeed be an important indicator for quality of sleep, but only focusing on improving this sentiment, ignores the other factors that can negatively influence quality of sleep (such as sentiment while falling asleep or poor sleep hygiene in general).
Table 8 Mentions of effective solutions to poor sleep related to
Sleep Time or Sleep As Android per narrative account
Bio Alarm
Clock Raising Awareness on Sleep Hygiene
Restriction of
Sleep Time Music Therapy
Interview Ex-User Sleep
As Android (User A) 1 1 Interview Frequent User
Sleep As Android (User B) 1 1 1 Interview Frequent User
SleepTime (User C)
1 1
Interview Novel User Sleep As Android (User D)
1 Interview Novel User
Sleep As Android (User E) 1 1 Interview Novel User
Sleep Time (User F)
1 1
Totals 2 6 2 2
An indirect advantage of using Sleep Time and Sleep As Andoid is something all Users agree on: a raised awareness on sleep hygiene. Using one of the applications seems to make users more aware of their sleeping patterns and habits in a way that reaches beyond the graphs of both applications. Adhering to sleep hygiene rules in general is mentioned often as beneficial to quality of sleep in all narrative accounts.
On their websites, raising awareness on sleep hygiene is not explicitly mentioned as one of the solutions of the two applications. However, all Users believe that using the application helped them become more aware about their sleeping patterns and the things that affect them. User C, for example, uses the application to look back at his sleeping patterns: “What does a month look like and is there a pattern to be found? Did I go out until late that one Thursday? What influence did that have on the days thereafter?”. User E describes it as such: “I think the most interesting thing it did to me is that it gave me sort of a framework to think about my sleep patterns. So having to use the application, made me think about it. So, even if I didn’t believe or trust those results, it made think about how I felt in the morning and what could have been the thing that led to this feeling. […] So I think it’s more about the awareness that comes with the use of the application, for me at least, and not the application itself. My sleeping habit seemed to become less random.”
The fact is, however, that both Sleep As Android as Sleep Time do not explicitly focus on raising awareness on sleep hygiene. So although, they (perhaps even unintentionally) help make users think about their sleep hygiene, the applications fail to offer actual advice on what good sleep hygiene entails. This missed chance results in users having to resort to other resources or their own perception of what is good sleep hygiene and what is not. Adding advice on good sleep hygiene to their applications could help Sleep As Android and Sleep Time address more of the wicked problem that poor sleep is.
Another lack of the awareness on sleep hygiene raised by Sleep As Android and Sleep Time is the absence of longitudinal data. As User C exemplifies, using the application makes him trace back his actions during the last month and tries to correlate them with the measurements of Sleep Time. In no way do the applications try to incorporate longitudinal data from, for example, a calendar or a food diary. Offering an option to import longitudinal data, would allow the user to reflect of his or her sleeping patterns and sleep hygiene within a contextual framework. This conclusion links back to the conclusion at the end of section 4.1.2.
User B and E also believed that it was useful that the applications helped them restrict their time in bed. They both used the CAPTCHA feature of Sleep As Android, which claims to get you out of bed in time “with a little task which needs to be solved first before the alarm gets dismissed. You can choose from various different CAPTCHAs: simple math equation, finding an awake sheep or shaking you phone to stop the alarm. An ultimate CAPTCHA solution are QR code or NFC tag based CAPTCHAs. You can place a QR code or NFC tag in your bathroom or kitchen. The alarm than won't stop until you get out of bed and scan the code/tag.” 11 User E called it “maybe the only useful direct feature”: “I realized that when I turned it on it had some benefits on waking up quickly or at least quicker than usual.” User B also experienced benefits from these feature: “You can’t hit the snooze button: you really have to get up. That works really well.” While the CAPTCHA option to a certain extent has the potential of restricting time in bed more effectively than a regular alarm clock (as one has to properly wake up in order to switch of the alarm), the way in which Sleep As Android employs this option has its limitations. Due to a lack of contextual knowledge of the user’s sleeping needs, the time in bed the user is restricted to seems arbitrary. The restriction of sleep time is based on the user’s own selection of a time to be woken up by the alarm clock. So although restriction of sleep time is considered to be effective, the CAPTCHA-approach to this does not necessarily take into account the ideal time the user should spend in bed. Besides that, after the CAPTCHA task is performed, the application does not bother to measure whether the user goes back to bed afterwards.
4.3 Solution Space Convergence
What this exploration of the solution space to poor sleep has, firstly, shown is that there are numerous ways to try and solve the issue of poor sleep. Much like with the causes of poor sleep, there are some solutions that are mentioned more often by the narrative accounts, but each narrative account has its own unique perspective on the solutions to poor sleep. This occurrence can be best explained by the sixth characteristic of wicked problems [11, p. 164].
An interesting conclusion is that although both applications do seem to offer both implicit and explicit tools to improve sleep, no User said that it solved their problem completely. In the line of the second characteristic of wicked problems (which claims that wicked problems have no stopping rule), this is not surprising. At the same time, for some Users the solution offered for the issues of poor sleep worked well and for others it did not: “3. Solutions to wicked problems are not true-or-false, but good-or-bad.” [11, p.163].
Although none of the solutions are good or false, certain solutions could be so bad that they inflict harm to the user, eventually making the problem worse. If restriction of sleep time for example
11 https://sites.google.com/site/sleepasandroid/doc/alarm
is performed without proper consideration, a user needing more sleep than he sleeps on average might, through improper sleep restriction, further build up a sleep deficit. So, what the Sleep As Android offers as a solution, can potentially also aggravate the user’s issue with poor sleep. In this sense, applications dealing with wicked problems have a certain responsibility in offering the best possible solution; a responsibility, which is also referred to by Rittel & Webber’s tenth characteristic of wicked problems.
5. DISCUSSION
The purpose of this research was to exemplify the overarching scientific problem of a reductionist approach to solving wicked problems, through a paradigmatic case study of Sleep Time and Sleep As Android. In order to exemplify the problem, this paper tried to explore to what extent these two applications measure and solve the issue of poor sleep, by performing a narrative analysis on the two applications, four academic papers, three experts, and six users.
The narrative analysis showed a discrepancy between the measurements methods and solutions used by Sleep Time and Sleep As Android, and the generally preferred measurements methods and solutions by the narrative accounts. Where the two applications objectively base their measurements of sleep quality on detected movement of the user, most papers, experts and users seem to agree that objective measurements are merely useful in combination with subjective measurements (and preferably only after poor sleep has been determined subjectively). On top of that, there seems to be a general distrust towards the measurements performed on the basis of the detection of movement by an accelerometer. Broadening the applications’ measurements with subjective alternatives might increase credibility and offer some contextualization of the objective findings.
The main feature of both applications, the Bio Alarm Clock, focuses on a very limited aspect of the wicked problem (namely only sentiment after waking up) and thereby ignores the complexity of the issue of poor sleep. The main use of the Bio Alarm clock is mostly of a gimmicky nature in the sense that it attracts the attention of users by offering a simple (scientifically questionable) solution.
More interesting is the finding that all Users claimed to indirectly benefit from using one of the applications, as it raised awareness of their sleeping patterns and habits. The general agreement among several Papers, Experts, and all the Users seemed to be that becoming aware of one’s sleeping patterns and habits can increase the adherence to rules of sleep hygiene. Thereby, it has the potential of benefiting sleep quality. Raising awareness on sleep hygiene was one of the two solutions that were thought of to help with the solution of three out of the four types of poor sleep. Considering the fact that neither Sleep Time nor Sleep As Android focus on one type of poor sleep specifically, incorporating raising awareness on sleep hygiene as an explicit feature could be a smart choice. Expanding the solutions offered by the applications by explicitly addressing and offering advice on sleep hygiene might increase efficacy.
The conclusion is that the Bio Alarm Clock-feature can be a great way to attract curious users, but seems to fail to improve sleep quality, whereas raising awareness on sleep hygiene seems to be considered far more effective, but could be a harder feature to sell and is currently not a point of focus for Sleep As Android and Sleep Time.
An earlier mentioned application that does explicitly focus on raising awareness on sleep hygiene is Slaap Lekker. This application, commissioned by three health insurers, bases its measurements solely on subjective input by the user and gives general tips and tricks on how to improve sleep hygiene. Examples of this include tidying up one’s bedroom and reflecting on thoughts that obstruct sleep. In future research, it would be interesting to research whether this application is more effective in improving sleep quality than either Sleep As Android or Sleep Time.
At the same time, this narrative analysis showed how the issue of poor sleep links to the characteristics of wicked problems of Rittel & Webber [11]. Besides the fourth and fifth characteristic, all characteristics linked directly to the findings of the narrative analysis. The fourth [11, p. 163] and the fifth characteristic [11, p. 163] deal with the (long-term) consequential effects of the solutions to wicked problems. Due to the short time-span of this paper’s research, all Users were interviewed only once. In order to explore the long-term consequences of the solutions that the applications offered––and thereby test characteristic four and five––at least one other interview with all Users after a period of a couple of months would have to be performed.
Overall, the approach of this paper to exploring both problem and solution space of the wicked problem Sleep Time and Sleep As Android are targeting, could serve as the basis for the further development of an application that tries to target the problem in a fashion that acknowledges the wickedness of the problem poor sleep poses. This approach, borrowed from the concept of Design Thinking [9], offers a way to sift a wicked problem without only focusing on those aspects which are measurable and/or solvable through a quantitative (reductionist) approach. It leaves space for alternative and conflicting views, while also searching for the most optimal solution to a problem. This is not to say that quantitative measurement and solution are by nature bad; the claim of this paper is that they are more useful embedded in a qualitative context.
6. CONCLUSION
This paper showed that Experts and Users perceive the reductionist approach of Sleep Time and Sleep As Android to measuring and improving sleep quality with skepticism. The objective measurements based on movement are generally found to be inaccurate and the applications could potentially benefit from the addition of subjective measurements.
The narrative analysis brought forward that poor sleep is a wicked problem, which is hard to define, but also that some solutions (raising awareness on sleep hygiene and sleep restriction) to this wicked problem reach across different definitions of the problem. Sleep Time and Sleep As Android however only explicitly focus on a minor sub-problem of the overarching wicked problem. Implicitly, both applications do already help with improving sleep hygiene, so it would be interesting to see what effects these applications would have if they were to incorporate advice on the adherence to the rules of sleep hygiene as an explicit feature.
7. REFERENCES
[1] American Academy of Sleep Medicine 2001. The
international classification of sleep disorders, Revised:
Diagnostic and Coding Manual. American Academy of Sleep Medicine.
[2] Bruner, J. 1990. Acts of Meaning. Harvard University
Press.
[3] Bruner, J. 1986. Actual minds, possible worlds. Harvard
University Press.
[4] Bruner, J. 1991. The narrative construction of reality.
Critical inquiry. 18, Autumn 1991 (1991).
[5] Bruner, J. 1960. The Process of Education. Harvard
University Press.
[6] Descartes, R. 1998. The World and Other Writings.
Cambridge University Press.
[7] Harvey, A.G. et al. 2008. The subjective meaning of
sleep quality: a comparison of individuals with and without insomnia. Sleep. 31, 3 (Mar. 2008), 383–93.
[8] Kelly, J.M. et al. 2012. Recent developments in home
sleep-monitoring devices. ISRN neurology. 2012, (Jan. 2012), 768794.
[9] Lindberg, T. et al. 2011. Design Thinking: A Fruitful
Concept for IT Development? (2011), 3–18.
[10] Morozov, E. 2013. To Save Everything, Click Here.
Technology, solutionism and the urge to fix problems that don’t exist. Allen lane.
[11] Rittel, H. and Webber, M. 1973. Dilemmas in a General
Theory of Planning. Policy sciences. 4, December 1969 (1973), 155–169.
[12] Robson, C. 2011. Real World Research. Wiley.
[13] Roth, T. 2007. Insomnia: definition, prevalence, etiology,
and consequences. Journal of clinical sleep medicine : JCSM : official publication of the American Academy of Sleep Medicine. 3, 5 Suppl (Aug. 2007), S7–10.
[14] Van Straten, A. and Cuijpers, P. 2009. Self-help therapy
for insomnia: a meta-analysis. Sleep medicine reviews. 13, 1 (Feb. 2009), 61–71.
[15] Tschimmel, K. 2012. Design Thinking as an effective
Toolkit for Innovation. … of the XXIII ISPIM Conference: Action for Innovation: …. (2012).
[16] Tsoukas, H. and Hatch, M.J. 2001. Complex Thinking,
Complex Practice: The Case for a Narrative Approach to Organizational Complexity. Human Relations. 54, 8 (Aug. 2001), 979–1013.
