Interactive Dreams

Johanna Rochegude May 2015

A playful interaction between

“Dreamers” and “Wakers”

Interactive Dreams

Thesis-project

Interaction Design Master at K3 Malmö University

Supervisor: Jonas Löwgren Examiner: Simon Niedenthal

I. C

ONTENT

I.

II.

III. Introduction ... 5

Context ... 5

What motivates the work ... 5

IV. Research focus ... 7

Purpose ... 7 Objectives ... 7 Knowledge contributions ... 7 Research question ... 7 V.

Sleep in playful interactions ... 12

Position of the project ... 18

VI. Methodology ... 19

Phase 1: Experiments ... 19

Phase 2: Design exploration ... 21

Phase 3: Reflection ... 23

VII. Processes ... 24

Selecting participants ... 24

Ethical considerations ... 24

Phase 1: Experiments ... 24

Phase 2: Design exploration ... 29

VIII. Phase 3: Reflect ... 41

Conclusions ... 41 Discussion ... 42 Further work ... 42 IX. Acknowledgements ... 44 X.

II. A

BSTRACT

This thesis aim was to design a new form of playful interaction engaging dreaming and awake players. In the tested concept, “Wakers” were able to influence and interact with the dreams of “Dreamers” (with the help of BCI to detect their brainwaves, emotional states and REM phases) by applying external stimuli on the dreamer (somatosensory stimulation, specifically vibrations). In the concept, the dreamer was wearing “the stimuli pajamas”, which vibrated in different ways every time the waker would poke, stroke, shake “the ball”, a prototype displaying the emotional states, sleep stages and movements of the dreamer. Each time the waker would interact with the ball, feedback would be transmitted to the

vibrating pajamas, thus influencing the dream and state of the dreamer, which would then be transmitted back and displayed on the ball. A new playful experience was created using sleep as a necessary component.

The research was experiment-driven (with body-storming and lo-fi prototyping), and revealed touch to be a powerful and underexplored way to influence dreams. Furthermore, transmitting the emotional states of the dreamer to “the ball” helped render the abstract notion of someone else’s sleep tangible to the waker. The co-creation session organized revealed that the particular concept developed in the context of sleep was tied to interesting notions, such as bringing forward the relationships between the players, the unbalanced power relations, sensual play, abusive play and more. The concept sketches explored the design space around the main concept and shaped some of these different scenarios. All these contributions are aimed to be inspirational material for further research in the field.

III. I

NTRODUCTION

C

ONTEXT

A conscious mind is a mind with a self in it (Damasio, 2011). We are only fully conscious when our self comes to mind. Thus, it can be surprising that sleep is neurobiologically considered as a state of consciousness, when it is culturally accepted as a loss of said consciousness. To such an extent that sleep and death were compared and merged in mythology and art (Hypnos and Thanatos, Greek gods of sleep and death, and brothers). However what are dreams, if not a form of consciousness, even if irrational? We owe most of our myths, legends, and bestiaries to our dreams, because we were imagining things while believing we were awake.

Brief overview of sleep

A brief overview of the last 50 years of research on sleepreveals that it is a phenomenon still widely misunderstood (Lawton, 2003). Guesses still prevail, with few confirmations over its purpose and intrinsic mechanisms (Foster, 2013). We do not know that much about it: we ignored it for a long time, due to its apparent passiveness (Bower, 1999). Moreover, it is a domain largely dominated by medical, psychological and biological fields of research. In these settings, some researchers started to focus on a peculiar area of sleep: the sleep cycles and their different phases, particularly paradoxical sleep. Said stage of sleep is home to the strangest natural occurrences provided by the human brain: namely dreams, hypnagogic hallucinations, and the like (as well as a sub-category of dreams known as lucid dreams; where the sleeper is aware that he is dreaming, enabling him to control varied parameters occurring in said dream such as characters, settings and courses of action). This break-through in dream research was enabled by Michel Jouvet, who discovered the existence of the paradoxical sleep, and Stephen Laberge, who is well-known for his research on lucid dreams.

Aim

The aim of this thesis is to design a new form of playful interaction engaging dreaming and awake players alike. BCI (Brain-Computer Interaction) and external stimuli during sleep reveal themselves to be the most promising tools to achieve this particular purpose. As a result, this project will evolve simultaneously within at least two different fields of research: sleep research (with a focus on the peculiarities of dreams and lucid dreaming), and Brain-Computer interaction (BCI, a subsidiary branch of Human-Computer Interaction, with a focus on playful interactions for the purpose of this project).

W

HAT MOTIVATES THE WORK

This project is motivated by several parameters:

Under-researched

The chosen domain is under-researched. One of the early examples of playful BCI is Brain Ball, a research led by Sara Ilstedt Hjelm in 2000. This research led to the creation of a one-to-one table game, where each player tried to send the ball to their opponent’s goal to score, using only their meditation brainwaves. Other researchers followed her lead in the BCI field, by building slowly upon each other’s’

research. However the most prominent researches led in BCI with a ludic purpose have yet to include sleep as a parameter. There is hardly any research linking actively sleep and ludic brain-computer interaction in the way it is intended for this current thesis project. The closest is an academic paper on a prototype called the Dreamthrower (Kamal et al., 2012). This leads me to believe that the domain is widely fruitful and not overly-researched.

Viable

The BCI field takes its roots back into the early century (with Hans Berger in 1929), however the actual expense of the equipment required prevented the field to blossom until the 90’s, where the technological break-through enabled BCI to attract a wider crowd of researchers and to emerge across several research fields, design included (Wijayasekara & Manic, 2013). Now, the existence of low-cost headsets with professional accuracy regarding real-time reading and analysis of the EEG signals

(electroencephalographic signals), like The Neurosky Mindflex device and the Emotiv EPOC headset, is allowing this project to be technically feasible, albeit potentially uncomfortable for the person sleeping. Regarding the interaction aspect between a dreamer and an awake counterpart, it is also achievable. In his research, Stephen Laberge taught lucid dreamers to move their eyes in a particular sequence in their lucid dreams (Laberge et al., 1981), to communicate with their awake counterparts while being asleep. Another example is the one with the hackers, who used a Zeo headset to communicate in a similar manner, lucid dreamer to lucid dreamer, thus “sharing” a dream (Top Coder, 2012). While being rather extreme, these examples prove that even without diving into the lucid dreaming territory, simpler interactions should not be a problem (interactions that does not require a lucid dreamer, but could be performed with a “normal” dreamer). These simpler interactions can be possible by applying external stimuli on sleepers to influence their dreams. Here, BCI proves itself to be the most suited tool to extract raw data from the dreamer, and feed it back to the waker.

Relevant

Finally, the project is relevant. Other researchers started to explore this field (although the theme and execution were quite different), making it clear that it is an interesting subject (Kamal et al., 2012).On another note, the knowledge produced could improve our understanding of brain cognition, especially regarding interactions with dreams. It could also change our understanding of sleep as a purely restorative process: sleep could then be considered outside of its medical sphere, as something that could offer unexploited possibilities regarding interactions between persons. This project could help provide new interactions that have yet to be explored. Here the use of BCI is strategic, for the BCI field is growing right now, and sleep research could be approached from another angle to broaden its potentialities. On another note, it would be relevant for interaction designers wishing to explore this design space, for game companies seeking new means of game-play, for early adopters seeking new forms of interactions, and for dreamers seeking new experiences.

IV. R

ESEARCH FOCUS

P

URPOSE

Like previously mentioned, the aim of this thesis is to design a new form of playful interaction engaging dreaming and awake players alike in a form of interplay. BCI and external sleep stimuli reveal

themselves to be the most promising means to achieve this particular purpose. The choice to turn the interaction playful, as opposed to “basic”, lies in the will to dissociate sleep from its intimidating medical and biological background, in order to reveal it as a valid state to be explored and used in interaction design.

O

BJECTIVES

In this project, wakers have been able to influence and interact directly with the dreams of their sleeping counterparts (with the help of BCI to detect their brainwaves, mental states and REM phases) by applying to them different external stimuli; thus creating an entirely new playful experience using sleep as a necessary component. This study has been motivated by the fact that sleep is gaining “popularity” outside of the medical sphere, due to the rising of the quantified-self movement (with small devices allowing consumers to track their sleep and giving them a sense of control, like the Fitbit, the Jawbone Up, etc). In addition, the appearance of low-cost and accurate BCI technology is propitious to the exploration of sleep and dreams outside of the medical scope.

K

NOWLEDGE CONTRIBUTIONS

First and foremost, the primary contribution to the interaction design field is the main concept: a playful interaction engaging dreamers and wakers, tested and analyzed. The findings stemming from this concept contribute to the understanding of sleep outside of the medical and scientific fields, as a valid component of a playful interaction. The secondary knowledge contribution is composed of concept sketches (design proposals) exploring different parts of the design space and aiming to provide

inspirational material for other designer/researchers’ appropriation. These sketches cover different kinds of interactions: from dreamers to wakers (associating a state of mind to a trigger on screen, mapping of states of mind while dreaming, etc.), from lucid dreamers to awake counterparts (eyes, brainwaves, neuro-training...), and more global scenarios. Furthermore, this thesis offers a preliminary analysis of the influences of diverse external stimuli on dreams (from wakers to dreamers).

R

ESEARCH QUESTION

As a result, the research question could be summarized as such: how can we design a new form of

V. T

HEORETICAL

F

RAMEWORK

Sleep research is mainly dominated by the scientific, psychological and medical fields: their main purpose is to try and understand the inner mechanisms of sleep, and help improving it. As a result, a straightforward literature approaching a subject like playful interaction between dreaming and awake participants is nearly non-existent, especially one also pertaining to Interaction Design. Therefore I perused diversified literature, each leaning toward one aspect of my project, in order to stitch them together and create the necessary framework.

S

LEEP OVERVIEW

With a project dealing with sleep and dreams, one need to understand better what is sleep and the hidden mechanisms of such a phenomenon. However it is not as easy as it may seem to be: Lawton has reviewed in 2003 over 50 years’ worth of research in the field of sleep and dreams, to stress the fact that “relatively little is currently known with any great certainty”. We do not know for sure why we feel this imperious need to sleep, or why sleep deprivation has such a drastic impact on our health, or even why we dream (Bondke & Persson, 2014). However researchers, anthropologists, psychologists and others have been able to observe in details how sleep and dreaming occur, as well as pinpoint a number of occurrences tending to influence it.

A wide range of mechanisms and influences

Sleep may seem passive and uneventful, but it’s not: in fact, there is a wide range of occurrences happening right beneath the calm surface, influencing our sleep and our dreams. These occurrences can be biological, social, oneiric, medical, and more.

Our sleep is altered by circadian, ultradian and homeostatic processes (Billard, 2008). These processes have been explored and altered in research and design: playing with the duration of our sleep, delaying our internal clock to erase jet-lag, syncing rhythms between humans, etc. Sleep is also a cyclic

phenomenon. It goes through several stages each night, repeatedly (see figure 1). The last ten years, numerous smartphone applications analyzing these different stages were created. They wake the person up at the end of their REM stage (Rapid Eye Movement, in paradoxical sleep), which is the most favorable moment to wake up (the transition is softer, and persons awoken during this stage are supposed to feel more rested). They also quantify your sleep, number of awakenings, heartrate, breathing patterns, etc. Users have a feedback spurring them to improve their sleep (the lucid dreaming application provides tips on how to improve sleep, depending on the sleeper’s statistics).

Figure 1: Anatomy of sleep

Peer pressure alters our sleep as surely as biology. Sleep is exclusively monophasic (Sciences et vie, 2013) in our Western society. Sailors, night workers and servicemen ought to have a polyphasic rhythm due to their job, but this pattern makes it harder to fit in society where we sleep for 8 hours straight at night. The anthropologist Carol M. Worthman complained about the “golden-rules” of medicine regarding sleep, while nobody studied it in its “natural” environment. After it was done, it was discovered that the vast majority had a polyphasic sleep (see Figure 2). We discovered that some societies had entirely built their communities around sleep. In the Northern Dene Settlement (Canada's community) people sought spiritual knowledge through dreams or altered states of consciousness (Goulet & Miller, 2007). Here, digital evolution helped collect ethnological data exponentially, to better know our sleep in various contexts.

Figure 2: Different sleep rhythms. From one block of 8 hours, to 6 power-naps of 20min.

Sleep is also the den of uncontrollable cognitive occurrences, which partly alter it. Dreams,

hallucinations, nightmares are phantasmagorical appearances which make us mumble, cry, toss and turn. Voluntary lesion of parts of the brain suppress the natural paralysis occurring during dreams (Jouvet, 1992). We can then witness the dream of someone, because he starts to act them like a sleepwalker would. Applications like Shadow, an innovative alarm clock, emerge to incite people to write down their dreams. These applications later offer a time-line to the dreamer, helping him analyze and review his dreaming wanderings. Objects such as REMEE (a sleep mask with embedded LEDs) help

people to become lucid inside their own dreams by waking themselves up from the inside, not unlike Inception.

However, sleep is not only altered by sane occurrences: there are three types of sleep disorders (Hauri, 2005): neuro-psychiatric disorders, dissomnies (insomnia, jet-lag, narcolepsy, etc) and parasomnies (sleep-walking, sleep paralysis, hallucinations and nightmares, etc). The means to hold these disorders back are varied: psychotropics, therapies, polysomnographies, diets, changes of habits... With the connected era, solutions such as auto-diagnostic and mobile coaches are emerging, partly thanks to the wirelessness and reduction in size of the devices. They track sleep, sleep-quality, vital signs, prevent sleep apnea, etc. Brain-computer interaction is also developing with the arrival of new, cheaper EEG headsets able to read brainwaves, notably the ones emitted during sleep.

The sleep research has been proactive regarding the understanding of its characteristics, processes and influences, especially when it comes to the anatomy of the brain and the different electromagnetic impulses allowing the detection of paradoxical sleep (and thus dreams). The technological evolution allowed the proliferation of devices related to sleep, however few of them really focus on dreams and fewer reflect upon how to influence them.

Dreams

Among all these occurrences in sleep, dreams prove to be the most puzzling. The dreaming state of mind offers similarities with an alert mind: effectively, findings show that “episodic recollections of

dreaming and waking experience are more similar in their process qualities (e.g., particular cognitive and sensory) than REM and Non-REM sleep" (Kahan & Laberge, 2011). This closeness to the waking

state is the proverbial door to interact with a different state of mind.

There are different types of dream, classified by the way they are used, and their behavior (premonitory dreams, creative dreams, traumatic dreams, resolutary dreams, etc). Creative dreams are dreams exploited in daily life (Michalko, 2012). They are free from the usual rational thinking accompanying awake behavior. Thinking is altered, logic abolished, giving birth to ideas through unconventional ways. Premonitory dreams and oracles were highly expected dreams, sometimes provoked (with the use of specific drugs), to decide of the future of the person or group who sought the advice of the gods. Hypnagogic dreaming occurs at sleep onset and has a high incorporation rate of memories of the previous day. They are truncated dreams, slightly more than hallucinations. They are relatively normal compared to fully bloomed dreams occurring during the REM phase (Stickgold et al., 2001). These REM dreams occur mostly during the paradoxical phase of sleep, and are the most well-known type of dreams. Their narratives are usually bizarre, chaotic and complex. It is possible to observe a person in this state without electronic help: she will have her eyes moving wildly behind her eyelids. In contrast NREM dreams, which occur during light and deep sleep, are hardly detectable without EEG headsets (Stickgold et al., 2001).

Finally, one type of dream in particular deserves to be cited: the lucid dream. Lucidity in dreams is defined as such by Laberge and Holzinger “a specific dream state characterized by the dreamer’s

awareness of being in a dream and the ability to volitionally control its content.” (Laberge & verified,

1990). They are a natural virtual machine, a simulator of real life where every limit is abolished. Lucid dreaming enables us to explore and actively use our dreams. It can have therapeutic effects like curing nightmares and sleep paralysis (Larousse medical, 2009). However lucidity in dreams is not a monolithic

notion: it can be achieved at various degrees, ranging from complete self-awareness to awareness of happenings outside of the dream, to semi-awareness without any influence on the dream whatsoever (Noreika et al., 2010). Complete self-awareness is the panacea, and it enables the dreamer to communicate with persons who are awake and within range: a prearranged pattern of eye movement can be detected on an EOG (Erlacher et al., 2003). It was a system set-up in order to prove the existence of lucid dreaming. Inducing a lucid dream can be quite difficult, however it can be achieved with commitment and a long training. It can be trained by mnemonic techniques learned during the waking state. These techniques include reality checks like checking the clock, reading one page of a book, etc. Each of these dreams offers a different set of characteristics, but all of them can be influenced by external stimuli.

External stimuli and their impact on dreams

In scientific research, several papers have been published on the different external stimuli possible and their potential influence on dreams. First of all, there are two main categories of external stimuli applicable: pre-stimuli (when the dreamer is aware of the stimuli being applied) and stimuli applied directly during sleep, without the awareness of the dreamer. The effectiveness of the stimuli in the first category can be trickier to test, since the actual awareness of the dreamer can count a lot in the incorporation of the stimuli in the dream. The different stimuli available are mostly linked with the different human senses: vision, hearing, smell, touch, etc.

Hearing

A way of influencing the dream is to make the person lucid. Sounds have been combined with lights in order to reach this goal. “In order to help dreamers realize that they are dreaming, external stimuli given

during REM sleep have been applied (e.g., tape recordings of the phrase “This is a dream,” conditioned tactile stimuli, and light).” (Laberge & verified, 1990).

Vision

Vision has probably been the most thoroughly exploited sense in sleep research, in order to influence dreams. Pre-stimuli wise, looking at pictures with positive, neutral, or negative effects during 30 seconds right before falling asleep (Carpenter, 1988) already influences the mood of the dreams. In another study, the viewing of rather unpleasant movies right before heading to bed influences negatively the emotions in REM dreams (Stickgold et al., 2001). Sleep onsets dreams have also been influenced by playing Tetris and several other video games right before falling asleep (Stickgold et al., 2001). When it comes to stimuli applied directly during deep or REM sleep, sleep masks equipped with LEDs on the inside are very common. The LEDs are triggered during REM sleep and flash at regular intervals (Vos et al., 2009). The dreamer sees then the lights above his eyes, directly incorporated in the dream. These sleep masks are sometimes coupled with sounds and are supposed to trigger lucidity inside the dream. Sensory stimuli and light cues increase the probability of having lucid dreams (Laberge & Levitan, 1995).

Smell

The influence of smells during REM sleep has also been tested. A study has focused on air-dilution olfactometry with fifteen participants. Rotten egg and rose smells were used on the test-subjects, and analyzed to see if there were any correlations between the smells used and the content of the dreams.

The good smells had a positive influence on the content of the dream, whereas the bad ones affected it negatively. Smells seem to primarily affect the emotions of the dreamer and the mood of a dream, but they are not directly incorporated in the dream as “smells” (Schredl et al., 2009).

Touch

Somatosensory stimulation has been tested using blood-pressure cuffs above the knees of several gymnasts. Here the stimuli was pre-applied. Interestingly, the stimuli has been processed in REM sleep as such (Sauvageau et al., 1998). Regarding the amount of research papers focusing on this particular topic, somatosensory stimulation seem rather unexplored as a way of influencing dreams.

Other

Finally, other senses have been exploited, such as vestibular motion (the vestibular system is the sense of balance and spatial orientation) (Leslie & Ogilvie, 1996). They found out that the rocking motion of a hammock could influence dreams, and enhance or induce lucid dreaming. Electromagnetic stimulation applied directly on the scalp is another vestibular stimulation, and allow “very precise control of intensity

and duration of the stimulation as well as the fact that the intensity of the electrical stimulation can be individually adapted for each participant” (Noreika et al., 2010). Another study focused on the possible influence that body posture could have on dreams: the prone position is more likely to induce sleep paralysis, hallucinations and lucid dreaming (Yu, 2012).

S

LEEP IN PLAYFUL INTERACTIONS

Dreams are rather easily influenced by external stimuli, however all these studies lack a component needed in this project: the playfulness. Unfortunately, the same can be said regarding interaction design projects: the ones focusing on sleep are doing so in order to improve it, not playfully interact with it.

Sleep and play

First of all, we need to define what play is, what would play with sleep potentially look like, and why it is relevant to associate it with sleep. In 2014, Miguel Sicart described playing as a “form of understanding

what surrounds us and who we are, and a way of engaging with others. Play is a mode of being human.”

He suggests that play can be as dangerous, as destructive and as corrupting as any other form of “being”: play does not necessarily have to be fun. It is pleasurable, but this pleasure can be taken from hurt, offense, addiction, not only from happiness and enjoyment. As a result different types of play exist, such as dark play, which pushes the limits between what is play and what is not play. Another relevant type is abusive play, which aims to “forefront the particular human beings behind gameplay” (Wilson & Sicart, 2010). A known example is Desert bus (Figure 3, Penn & Teller, 1995) which is a game that requires the player to drive a bus to Las Vegas in real time, for 8 hours straight, and the game cannot be stopped. The player who finally reaches Las Vegas earns only one point. Another example of pleasure drawn from pain is “Hurt me plenty” by Robert Yang (Figure 4), where the player can spank a character using the leap motion. The player needs to be attentive to the character, and stop immediately if this one utters the safe word. It takes its roots back into the BDSM communities and try to display how they view intimacy and consent. Interacting with the dreams of someone might turn out to be a form of abusive play, at least for the person subjecting themselves to the role of the dreamer. It might also turn out to be a form of very intimate play, which can turn into a more sensual play.

Figure 3: Desert bus, by Penn & Teller, 1995.

Furthermore M. Sicart states that play is a way of freeing people from their moral conventions, while making them present and raising awareness regarding their weight and importance. Associating a playful behavior with sleep, especially with a number of participants as low and intimate as two, will certainly bring up notions like power relations. The dreamer is vulnerable and has to trust the waker with his sleeping body and his dreams. The notion of relationships will probably be brought up too: the forced intimacy between the dreamer and the waker could break social conventions, the relations between the participants being pushed into the foreground. These power relations might be accented by the obvious roleplaying taking place between the two participants. One is the dreamer, the other is the waker. In 2006 Caillois established a classification of games, with 4 main categories: Agon (competition), Alea (chance), Mimicry (simulation) and Ilinx (Vertigo). Clearly the playful interaction with dreams intended in this project belong to the vertigo category; with the possible thrill that being vulnerable might offer. These power relations will probably also need a set of rules, implicit or not, between the two participants.

Another notion relevant to this project is the difference that exists between play and playfulness: play is a contextual and disruptive activity. Playfulness is an attitude which, while lacking certain characteristics common to play, enables the player to appropriate and occupy the surrounding context, even if it was not intended for play (Sicart, 2014, Salen & Zimmerman, 2004). Associating play with sleep is a way of dissociating it from the medical and research sphere. A way of extracting it away from its known boundaries, to explore it from another point of view. It will enable the player (and designers stumbling upon the thesis) to reconsider the situation of sleep and dreams, thus opening doors to generate a dialogue where new designs and new knowledge can emerge from.

Rendering dreams interactive in a playful interaction between dreamers and wakers would not have been possible a few years ago, with the lack of affordable technologies able to capture accurately the different electromagnetic impulses and brainwaves of the dreamer. However the retrieval of the needed data to elaborate such interaction is no longer an issue, not with the rise of the BCI field and its growing expertise.

Ludic Brain-Computer Interaction

If playfulness is not associated with sleep research and dreams, it is fortunately present in the different artefacts produced in the BCI research, and could offer a starting point to later move on toward

playfulness with dreams. The BCI field is an emerging research area of Human-to-Computer Interaction (HCI) currently providing promising means of interaction outside of the pool of standard devices: effectively, its main goal is to use brain activity alone in order to interact with systems, bypassing physical motor activities (Wolpaw et al., 2002). A great number of works within the BCI community are focused on the medical side (Kaufmann et al., 2012), taking advantage of the revolutionary improvement it could make regarding the quality of life of certain people (how to help paralytic persons control wheelchairs or robotic prosthetic with their own thoughts, etc).

A literature overview of the BCI field proved it to focus mostly on the genesis of the research field, the scientific understanding of the brain and its different waves and the existing technologies to detect these waves. While some of the papers are a general survey and assessment of the BCI field for the last decades and its different purposes, Ferreira et al., (2013) offer a complete analysis of the BCI field and its diverse expressions, and gives a review of all BCI prototypes in gaming to this day. Some of them will be more thoroughly addressed later. What is interesting for this thesis is that this research is focused on

BCI for healthy users, seizing its potential as a new form of game-play control within the ludic interaction realm (Coulton et al., 2011). However most of these works involve a single player with the computer, and take advantage of only one or two unique interaction possibilities (using mainly attention and meditation levels). However, the data retrieved is far more substantial than these 2 outputs: delta, alpha, gamma, beta, theta brainwaves could be collected, as well as the emotional states, clenching of the jaw, twitching of the eyes, furrowing of the brows. All this data could lead to richer interactions if it was to be exploited. Furthermore, it can be collected in sleep, and help differentiate different sleep stages too (like light sleep, deep sleep, paradoxical sleep and when a person is dreaming). However regarding sleep, as expected, most of the BCI technology presented in these examples has been used with the purpose of inducing lucid dreaming, with no gaming component added whatsoever, and no use of the interaction potential current EEGs offer in terms of data retrieving.

Related works & relevant design exemplars

As stated above, here are several examples worth mentioning. In “analogic” games, the closest we have to interactions involving sleep and play is “Werewolf”. It is a game where some players are werewolves, while the rest are villagers. Each “night”, the werewolves shift and kill a villager, while everyone else has their eyes closed, feigning sleep. Then, everyone wakes up, the dead is announced, and villagers debate on who might be the killers before agreeing on eliminating one suspect. The game goes on until there is only the werewolves or the villagers remaining. This game is worth mentioning for the

active/passive mechanics that it displays (even if players do not actually fall asleep), which is similar to the one that will be enacted in this project: some players are active at night while the others are passive, vulnerable and at their mercy while sleeping.

Brainball

One of the pioneer work using BCI is the Brain Ball, a research led by Sara Ilstedt Hjelm in 2000 (Figure 5). This research led to the creation of a one-to-one table game, where each player had to send the ball to their opponent’s goal, using only their brainwaves. Entire control of one’s state of mind was required to win. What is most interesting in this work is its gameplay mechanics: in order to win the game, the player has to be (seemingly against all gaming traditions) as relaxed and unfocused as possible, even nearly on the verge of falling asleep. It makes this project one of the closest to linking BCI, play and sleep-related states of mind.

Figure 5: Brain Ball, by Sara Ilstedt Hjelm, 2000. Picture from http://origin.arstechnica.com Teegi

TEEGI is a tangible electroencephalographic interface built by Frey, Gervais, Fleck, Lotte and Hachet (Frey & al, 2014). The prototype was presented in a paper published in 2014. Succinctly, Teegi is a small humanoid object where the brainwaves of the user are displayed in real time, with the help of augmented reality (Figure 5). Its main purpose is to offer a tangible and intuitive way to learn about the brainwaves and their associated behaviors, through playful and tangible feedback. The humanoid comes with other smaller figurines, acting as filters to select which brainwaves are to be displayed on the main humanoid. They explain in a quite detailed way the BCI field and the evolution of the EEG, as is accurately described the technology used in the prototype, allowing others to build upon their research. While not having any close ties to the sleep research field, this project is interesting in its will to bare the mechanisms of the brain to its user. The brainwaves’ new-found tangibility makes them more relatable and less seem like abstract concepts with no grounding in reality. Regarding sleep, it brings up an interesting question: would a person have a better grasp of something as intangible as sleep or a dream, if one were to have a tangible representation of it and its mechanisms, in the palm of their hand?

Figure 6: Teegi, by Frey, Gervais, Fleck, Lotte and Hachet, 2014.

Admiral Mind Battleship

This paper presents a thorough compilation of all BCI research games up to this date (Ferreira et al., 2014). They talk about projects like the Finding Star game developed by (Ko et al., 2009), where the user is controlling a character using the standard controls (keyboard/mouse). However their degree of concentration can influence the rendering of the game and make it unstable. When the character needs to recover, the player has to be relaxed. Other games exploit the potential offered by EEG headsets, like the recognition of emotions. (Kang et al., 2012) designed a horror puzzle game, where “screamers” appear randomly, terrorizing the player. In this context BCI allows to catch this fear, and exploit it. Similar to this game, Basori developed in 2013 a project called the Emotion Walking, an EEG headset coupled with a glove to detect the emotions of the user and display them visually onto an avatar. Finally, the Painting Interaction (Huang & Lioret, 2013) takes advantage of the headset’s ability to detect “thought-movements” (you can train the headset to recognize the action “push”, when you think about

pushing, and then link this action to an external software). Here, the person can think “paint”, and a painting will start to appear, as if the person was controlling the brush strokes with his brain. Dreamthrower

Finally, the Dreamthrower is a project by Noreen Kamal, Abir Al Haira, and Sidney Fels (Kamal et al., 2012). This project explores how to “create, throw and catch” dreams. Technically, it is a device that detects the REM stages of the user, before applying a sound and light stimuli to influence the dream. After having the dreams, users can report them on a special social network, coupled with the stimuli that they used to influence said dreams. The idea behind the social network is that you can “throw” the stimuli and dream to another person, in order for them to experience said dream on their own, hence the “Dreamthrower” name. It is more of a collaborative “gaming” experience.

Some projects outside the research sphere are also worth mentioning for their incorporation of sleep (most projects tend to be about sleep though, using it as a theme rather than an actual playful component). There is the experience of the two hackers who “shared” a lucid dream using a Zeo headset (Top coder, 2012). A few applications are designed to help induce lucid dreaming, and help themselves by tracking your motion in your sleep, thus detecting your sleep stages (like the Shadow application which works primarily with voice notes, the Lucid Dreaming App, the sleep mask Remee, etc.). An RPG was also developed to be used as a patch on top of the Fitbit application, using the data (how many miles walked, how well you sleep at night) to complete quests and go on adventures.

P

OSITION OF THE PROJECT

It has been very difficult to find research papers or examples dealing with all the parameters that are implemented for this study, considering the chosen problem domain. Most of the research describes single and iterated experiments ranging from playful interaction to actual thought-through games using EEG in the context of BCI (Ferreira et al., 2014). However only one of these experiments focuses on sleep and dreams and how it could take part in an actual game or playful interaction. As a result, the current project could be considered as a rather new ramification of BCI research, focusing only on sleep-related states of mind and their potential use (here in playful interactions).

VI. M

ETHODOLOGY

To create a new form of playful interaction between awake and dreaming players, I propose the following methodology.

P

HASE

1:

E

XPERIMENTS

The first step for this particular project is to conduct experiments. These experiments constitute the observational phase of the project.

This phase is needed to familiarize oneself to what the brainwaves look like under the different EEG scanners available, and what each one of them has to offer in terms of potential interactions. Although valuable, these insights are not mandatory: a great amount of scientific papers in sleep research have been extensively covering the sleeping brain and its different brainwaves with great accuracy, more than what will be needed for this project. They will be observed here not for accurate research purposes, but as inspiration to be used for further development of the interaction concepts during the design

exploration phase.

This phase is also needed to familiarize oneself with the different kinds of external stimuli applicable to a dream, and how each one of them influences differently said dream. These tests have already been performed extensively and with great laboratory control in the scientific community (see chapter V: theoretical framework). Once again here, the accuracy and strict protocol control of these experiments are not a priority, for they primarily serve the purpose of familiarizing oneself with the technology and the interaction potentialities for further design concepts exploration.

Sleep self-study

The first need of the observational phase is to learn how to use the tools at disposal (This learning will be performed throughout the entire project). These tools are for now : the Epoc Headset, capable of reading basic thoughts (if trained beforehand), emotional states, facial expressions and EEG outputs, the Mindflex EEG headset, capable of reading 10 different channels of brainwaves (alpha, beta, delta, theta, gamma, etc), the Mindwave headset, capable of the same thing but with additional software and mobile-friendly, and other mobile applications offering sleep-tracking methods in a sufficient manner (such as sleep as android, lucid dreaming app, etc.). These tests and calibrations will be performed on myself.

Stimuli exploration 1

The second need of the observational phase resides in the testing of the different stimuli able to influence dreams.

Goal of experimenting stimuli

The main goal of these stimuli experiments is to determine which stimuli has which effects, and which one has the most impact and is the most likely to be incorporated into a dream. This stimuli will then be considered the most promising for further interaction research.

How: main

For this, one participant at a time will be observed. More would be too complicated for just one observer. Once the participant enters a REM stage, the observer will perform the stimuli themselves (it can be visual or audio cues, somatosensory or olfactive stimuli, as brought up in the theoretical framework). The sleeper will then be woken up some time after the stimuli was applied, to measure its impact on the dream. Here the project relies heavily on the participation of external people willing to let their

paradoxical sleep be experimented with. Effectively, it requires of the participants to spend the night under my scrutiny, falling asleep, be stimuli tested and then awakened during their dream phase to account for what happened in the dream and what might have influenced it. Each of these long

observational nights will report 2 to 3 stimuli testing (One per REM cycle, a cycle happening 2 to 5 times a night depending on the sleeper). Furthermore each stimuli test will result in a semi-conducted

interview during the night, so as to not risk losing the contents of the dream to memory failure. Each night will, if possible, result in a debriefing in the morning with the sleeper. Due to the rather heavy settings of these experiments, it is preferred to perform a qualitative testing, with fewer persons (maximum 2-3) but on a longer period of time, so the participants get accustomed to the observer and their presence to reduce eventual interferences.

These experiments will scaffold as such:

How: dream diaries

Stimuli experiments can be implemented without the need for a headset. However to prove successful, all willing participants will need to sustain a dream diary (at least in the beginning) in order to improve the recollection of their dreams the morning following the observation session. If participants do not remember their dreams, the observation phase will be greatly hindered.

How: semi-structured interviews

The experiment will have a semi-conducted interview each time the participant is woken up (on each REM stage, so around 2-5 times a night). The participants, their feedback and dream recollections will remain anonymous. To determine which stimuli or which combination is the most efficient, several questions will need answers: did the person notice a difference between this dream and the usual ones? If yes, how did the stimuli influence the dream? Did it change the overall atmosphere of the dream? Was it directly incorporated in the dream? Did the stimuli provoke a lucid dream? It needs to be noted that the presence of the observer will probably influence the results, just by assuming their role next to the sleeper.

How: qualitative research

The pool of “test-subjects” will be limited to a two persons’ study. Effectively, due to the nature of the project (on sleep, resulting in invasive and time-consuming experiments), the methodology must favor a qualitative research: participants are required to be observed for long periods of time, and it involves them sleeping and dreaming. These instances are subject to tricky conditions that are not always controllable: testers have to establish a trusting relationship with the observer, they have to get used enough to the situation to fall asleep with minimal disturbances, they need to remember their dreams and the observer must be stealthy enough to apply stimuli without waking the testers. As a result, the project presents a hard learning curve for the participants involved in the project (mainly the test

subjects) and the observer. Furthermore the use of sophisticated technology prevents a large number of testers. As a consequence, the design process will reflect these choices and the project will present traits that are seemingly the opposite of traditional Human-Computer Interaction: it will present a high learning curve, quite similar to instrument making, with a very high threshold and expectations levels. How: evaluation criteria

All stimuli will, if possible, be tested on both persons to cross-check results, even if it is obviously not possible to draw general conclusions from such a low pool of participants and might of course not apply to other people in the same conditions for the same tests. Pre-stimuli (stimuli applied before sleep, with the awareness of the tester) and stimuli applied directly during sleep (without the awareness of the tester) will be tested. A session will be a pass if it can be determined that the dream has successfully been influenced by the stimuli applied beforehand.

To determine if a dream has been successfully influenced, certain things must be taken into consideration. The dreams’ results might be biased by the followings:

- What is the person usually dreaming about?

- What did the person do the day of the experiment before going to bed? - Was it a pre-stimuli or a stimuli applied directly in sleep?

Results have to be carefully considered and cannot be generalized. They have not been realized within a medical field and have not a scientific purpose.

Ethical considerations

Ethically, it needs trust and the guarantee that the data retrieved will not be leaked, and that the stimuli applied will respect the person’s integrity. Everything that has been done in the frame of this project has been done so with the informed consent of the participants: there is no photographs or recordings of the experiments, and the transcript material has been modified to respect the persons’ wishes.

P

HASE

2:

D

ESIGN EXPLORATION

The second step of the methodology for this particular project is the design exploration phase. This phase will build upon the findings of the previous experiments. It will narrow down the possibilities regarding how to apply the chosen stimuli and it will focus on researching different interaction design concepts.

Stimuli exploration 2

Once an agreement is reached regarding which stimuli would be best suited for the interaction concepts, this stimuli will be further investigated and several variations of it will be tested, with a clear focus on how to adapt it to a design concept. The techniques employed here will be the same as the ones used in the first stimuli exploration.

Concept sketching

Concept sketching is a tool that helps lay down ideas quickly, while retaining the necessary technical aspects that render the imagined interactions plausible. These sketches focus on technical needs for the

concepts, and lay out scenarios for the kind of effects I want to achieve, for both the dreamer and the waker. They will be held simultaneously with the second stimuli exploration, and will aim to find a suitable concept that will later be prototyped. Beyond this, the aim of these concept sketches (design proposals) is to constitute a valid secondary knowledge contribution (Gaver & Martin, 2000), along with the main concept developed. Described by Gaver, the overriding function of design proposals is to “serve as landmarks opening a space of design possibilities for future work. As such, the concepts are

placeholders, occupying points in the design space without necessarily being the best devices to populate it […] in this way, the proposals acted as probes into our values and beliefs, eliciting a conversation about the directions we might take in pursuing the research. […] Beyond serving as suggestions for development, then, design proposals can also be seen as complex hypothetical statements for debate.”

Body storming and co-creation

The aim is to have a concept of a playful interaction where someone is dreaming and an awake counterpart is alerted of their status; they will in return offer a stimuli, triggering a response from the sleeping person. This concept (or this staging of the concept) will be user-tested and the session documented.

The actual prototyping process will take the form of a collaborative exercise, with at least 2 co-creation sessions. Methodologically, these sessions will be divided into three parts:

- Phase 1: The dreamer is falling asleep. I will explain the concept to the waker and ask them to “play pretend”: body storming around the concept and act it out. All the while asking questions about their feelings, etc.

- Phase 2: I will wake the dreamer up, and ask them questions on their dreams, how they feel, on the stimuli, etc.

- Phase 3: I will have a general debriefing with both dreamer and waker, on their general thoughts on the concept, the co-creation session and how to improve/change the experience. In a nutshell, I will be the mediator between the waker and the dreamer. I will explain what I will roughly do, what is supposed to happen, and then start a debriefing session with all participants at the end of the experience, to refine the concepts, detect mistakes and possible improvements. Such setting is very similar to the Wizard of Oz method (Dow et al., 2005): I will be the wizard operator, simulating the entire interaction and the feedback loop between the dreamer and the waker. As stated by Dow, WOZ technique exists to explore the design space without having to implement time-consuming and challenging hardware, thus allowing the designer to quickly evolve in the design process and focus on the user feedback, without technical impediment. The design is evaluated before being actually built.

User-tests

User-tests will be a part of the co-creation process and will provide valuable feedback that will help shape the final concept up. This feedback will help pinpoint the most efficient way to translate the dream and emotional state (or communication if lucid) of the dreamer into efficient feedback for the persons awake, and vice-versa, and how to make the experience exciting for both participants.

Expected results

Ideally, this project is intended to produce knowledge actionable inside and outside the field of interaction design research. Inside the interaction design community, the knowledge expected to be produced is first and foremost a tested concept which would hopefully lead to other projects including dreams and sleep-related states of mind as valid interaction material. It could prove to be valuable to the community in the sense that it might shed a new light on how to address sleep and how states of mind and paradoxical sleep in particular can act and be used as controllers and valid interaction components, thus opening new doors on how to envision playing and interaction itself with objects, depending on the state of mind of the user. Part of the knowledge is expected to take the form of concept sketches (design proposals) as described earlier. Finally, a great part of the expected knowledge is purely analysis, with the observation sessions of the impact of diverse external stimuli on dreams. These sessions would appropriate methods from the field of sleep research and adapt them for interaction design purposes. These methods would lead to propositions of different possible interactions between lucid dreamers and awake counterparts, and simple dreamers and awake counterparts. This knowledge would be valuable as it starts exploring the diverse possible ways of interaction between sleepers and their awake counterparts.

Outside of the interaction design community, these knowledge contributions might have a societal impact on the way to apprehend sleep and dreams altogether: not as a mysterious regenerative power anymore, but as something that could be used, playfully. It would redefine our conception that sleeping people are unreachable and inactive.

P

HASE

3:

R

EFLECTION

The third phase will compile the findings and analyze them. It is the conclusion. These findings, together with the concept sketches, will be aimed toward the design community regarding the introduction of sleep and dreams as valid entities that can be used as components of potential interactions.

VII. P

ROCESSES

As stated in the previous chapters, the aim of the project is to design a playful experience where awake people (the wakers, for a lack of a better term to pinpoint the role of the alert counterpart) can easily detect when a sleeping person is in a dreaming state and apply an external stimuli to influence the dream. The reactions of the dreamer to the stimuli are then transmitted back to the waker, in a feedback loop.

S

ELECTING PARTICIPANTS

Influencing the dreams of a person is not an easy feat: it requires a number of conditions which gathered together greatly improve the odds of success. One of these conditions is the careful selection of the participants. Effectively, it is crucial for them to have a natural tendency to recall their dreams, so as to be able to remember and speak about them once awoken. They also need to be able to easily fall asleep anywhere so as to be hardly disturbed by their surroundings (namely the observer taking notes, typing, lights in the room, stimuli applied). Finally, a trusting relationship needs to be established between the observer and the participant.

E

THICAL CONSIDERATIONS

By trying to influence the sleeper’s dreams and interact with their general mental states, we are closing in on taboos such as mind-reading, and manipulation of sleep. Although we’re still far from such technology being operational, it still raises inquiries that have to be kept in mind regarding concepts like privacy, morality, and the potential vulnerability of the sleeping person. However the participants were educated on the nature of the experiments and were entirely willing. They considered the experiences to be thrillingly fun, even if some of them might seem rather “violent” for the outside world. As a result, proper psychological safeguards is needed for further research, as is the enlightened consent from the testers. On a side-note, no pictures of the testers have been taken during the experiments, to respect their privacy.

P

HASE

1:

E

XPERIMENTS

Learning the tools

Before starting the stimuli experiments, I performed a series of tests using varied technologies, to grasp the mechanisms of sleep. From there, I had a better understanding of the gathered data, and the possible ways to exploit it in an interaction scenario.

I tested different EEG headsets available on the market. Briefly, the Epoc headset was able to track movements, facial expressions, brainwaves, mental states, and to associate “actions” to these mental states (as done by Huang & Lioret, 2013). However to achieve this level of precision, the device required an extensive cognitive training from the user to associate their brain patterns to the actions.

The headset also needed to be handled with care, being fragile and ill-fit to be used for sleeping (Figure 7). The Mindwave, another headset of inferior standard, could track the brainwaves, the attention and relaxation threshold, and blinking (the equivalent of this headset was used in BCI games like the Brain ball). This device could also be worn in bed, if careful (Figure 8). Both devices were accurate in their measure, and could be hooked up to a third party software to detect sleep stages (such as Lucid Scribe). In terms of feasibility for the final design concept, the Insight headset appear to combine the wearable easiness of the Mindwave with the professional data collection of the Epoc+ (allowing a wider range of possible interactions, Figure 9). (Appendix 1: Headset testing).

Figure 7: Epoc headset, Emotiv.

Figure 9: Insight headset, Emotiv.

In parallel, I asked my soon-to-be testers to keep a dream diary in order to prepare for the stimuli experiments. They kept it for a week in order to train their brains to recall all dreams that may have occurred throughout the night (Carpenter, 1988). On a side note, lucidity is better achieved if the dreamer has trained his memory for the recollection of his dreams. The more you remember your dreams, the more you are familiar with your “dreamscape” the more you can recognize you actually are in one (Noreika et al., 2010). Each entry of the diary were composed as such: the nature of the dream, the personal reactions/emotions to the dream, what could have influenced the dream before going to bed (TV show, music, etc).

Finally, I tested diverse applications to track my sleep: Fitbit (a pedometer that also tracks sleep, which proved to be very inaccurate), smart alarms (tracking sleep in order to wake you up at the best possible time), and finally an application called Lucid Dreaming, which proved to be really technical and really thorough (Figure 10 & 11). Designed to induce lucid dreaming, it used the accelerometer on the phone to pick up movement on the bed and thus tracked down which sleep stage I was in. From there, it triggered light and sound cues to make me realize I was dreaming. I confirmed here that sleep was divided between several cycles, which repeated themselves 3 to 5 times each night. Each cycle were roughly 90-100 minutes in length, and went through different sleep stages: light sleep, slow-wave sleep, deep sleep, and finally paradoxical sleep (chapter V: sleep overview). In paradoxical sleep, REM happened, as well as micro awakenings, before entering another sleep cycle. The REM phase is generally very noticeable because the eyes of the person are moving wildly, and the face becomes very expressive. It can occasionally be accompanied by sounds and mumbling (Stickgold et al., 2001). Interestingly, paradoxical sleep occurs in greater quantity in the last hours of the night, which might be why we seem to more easily remember our dreams in the early mornings. These tests gave me a foundation to build my stimuli experiment: wait around 90 minutes for the person to fall asleep, check periodically the eyes and the face, as well as eventual sounds they might emit and only then, apply the stimuli.

Figure 10: Lucid Dreaming app, sleep tracking.

Figure 11: Lucid Dreaming app, advices.

Stimuli exploration 1

For this first phase of experiments, 2 participants agreed to be observed at night. The experiments that were conducted were low-tech: I sat down next to the sleeping persons, and used myself as a “home-made” REM detector (WOZ, Dow et al., 2005). It prevented me from worrying with the time-consuming elaboration of a prototype using an EEG headset, and thus allowed me to spend more nights at the bedside of people. I waited for them to start dreaming, before testing different stimuli. I tested the efficiency of different stimuli delivered before and during dreaming. This first phase of experiments was a wide stimuli exploration (light, touch, sound, etc), which purpose was, in conjuncture with the readings, to determine which stimuli would be best fitted for a more elaborate interaction concept.

Findings

Experiment n°1

Person A was observed Thursday 19 until 3 in the morning through a Skype call.

She went through one REM stage, during which she was visually and aurally stimulated by this particular website (staggeringbeauty.com). This was made possible by sharing my screen on her computer, enabling me to influence her. After her REM stage, she was woken up and asked to recall her dream. Although she was not able to remember what had happened, she distinctly remembered a “really harsh change in rhythm”. The dream was normal and then “it wasn’t”. At the end of the stimulation, she knew that the sound was external, but she kept dreaming and did not wake up. The combination of light and sound thus triggered some degree of awareness of her surroundings, without evolving into a fully grown lucid dream (see different stages of lucidity, Noreika et al., 2010). The stimuli had distinctly influenced the dream.

Experiment n°2

Person B was observed Friday 20 until 5 in the morning, in her bedroom.

She went through 2 REM stages. During the first one, she underwent somatosensory stimulation (since it was the first experience with this person, I started with a pre-stimuli to put her at ease). She went to sleep fully aware of her ankles bound together by a soft piece of clothing (similar to the gymnasts in Sauvageau’s paper, 1998). After the REM stage, she was woken up and asked to recall her dreams. Interestingly, the stimuli was directly incorporated within the dream, along with the observer, who “tortured” the sleeper and forced her to walk around with a cable around her ankles. The sleeper was begging me in her dream to “take it off”, she felt like it was cutting her blood circulation and that she was going to faint. She then rushed into a bar to borrow scissors to cut the rope, but it turned into hair. The somatosensory stimulation proved to be very effective and directly incorporated into the dream. Part of it may be due to the fact that this was the first experience with this person, that the stimuli was applied when the person was awake, thus building anticipation and excitement on what would be the result of the experiment.

During the second REM period she was stimulated by the smell of vinegar. Here, the stimuli was not directly incorporated into the dream (it mostly influenced the mood of the dream, like in the experiment of Schredl et al., 2009). Although the dream had a distinct morbidity into it and something was

“bothering her”, we could not really pinpoint if the vinegar was at fault. In the morning, a debriefing was done regarding the events of the night. The tester was really enthusiastic and surprised by how powerful the somatosensory stimuli was.

Next experiments

Before beginning this first phase, I thought about testing stimuli on a pool of 4-5 people. In light of these first two experiments (time-consuming and energy taxing), I decided to focus on only 2 persons (as stated in the methodology), but for a longer duration. This allowed me to detect the existence of a learning curve in the participants and how they adapted to this new situation.

In regard of these preliminary experiments and the literature perused, I decided that the next experiments were going to focus on the different interactions possible with the dreamer using touch. Effectively, even though it seems really promising, few papers discuss of somatosensory stimulation to

influence dreams (Sauvageau et al., 1998). All “design interactions” with dreams are focusing on lights or sounds (mainly to induce lucid dreaming, see vision part of theoretical framework). This second wave of experiments were going to help decide what kind of somatosensory stimulation is the most efficient, and how slight modulations could differently affect dreams. Each of these experiments were going to help conceive an interaction concept.

P

HASE

2:

D

ESIGN EXPLORATION

Figure 12: Feedback Loop from waker to dreamer, enumerating different ways to interact.

Body-storming

The next performed experiments followed the same patterns as the previous ones: observe a sleeping person, then apply a stimuli when the person is dreaming to influence the dreamscape. Although in appearance these experiments may seem to be more related to scientific research on sleep than to interaction design, I fully consider them to be design experiments: each of them were, on their own, a very lo-fi prototype (or a body-storming session) designed to explore the design space, step by step, and to refine the interaction from waker to dreamer (Figure 12, upper part). Each experiment were built upon the conclusions of the former, and the result was later integrated into a grander scheme.

Findings

Experiment 3

Since it was her first experiment, the stimuli applied was a pre-stimuli, taking the form of a strip tightly wrapped around her left forearm (Figure 13). Once again, the pre-somatosensory stimulation proved to be really efficient. The stimuli was incorporated in the dream, but not directly though: the person felt like her entire left side was paralyzed and she had to drag it around. Again, my presence as the observer was negatively perceived, as I was again incorporated in the dream as a villain, chasing the

handicapped dreamer around. Given the efficiency of this stimuli, I tested it again in experiment 5, but this time directly in sleep, so the dreamer had no way of “self-influencing” themselves by knowing in advance the stimuli.

Figure 13: Somatosensory stimulation with bind Experiment 4

Person B was monitored April 3rd until 6 in the morning, in her bedroom.

Here the person went through 2 monitored REM phases. Another type of somatosensory stimulation was applied, voluntarily less invasive and “violent” to see if the use of another medium would differently influence the dream. Wet wipes were applied to the arm, along with tickles (Figure 14). The person showed signs of being aware of the stimuli, but kept on dreaming, without really reacting or including it in the dream. It is perfectly possible that I caught the dreamer too late in the REM phase (since she had a micro-awakening before turning and going back to sleep) or that the stimuli was not strong enough.

Figure 14: Somatosensory stimulation with wet wipes

Experiment 5

Person C was monitored April 13th until 3 in the morning, in her bedroom.

Here the person went through 2 monitored REM phases. To build on top of the results of the previous experiments, I tried tying a strap of clothing around her arm when she started dreaming. Touch can be powerful, especially if it’s a pre-stimuli. However directly applied during sleep, both persons (referring also to the previous experiment) felt the stimuli, woke up, or were on the verge of waking up when I applied pressure directly during both REM sleep and deep sleep.

For her second sleep cycle, I tied the cloth beforehand on her arm and snuck in a pedometer with an alarm set to vibrate during the next REM phase (Figure 15). She was not aware of the way the

pedometer would vibrate. When I triggered the device, her dream recollection was a nice chat between friends when the buzzer signifying the food was ready started to make the table vibrate. This way, the testing of the stimuli was successful, while minimizing the chance of waking the dreamer up by the movements of the observer. Interestingly, even though I was once again incorporated in the dream, I was not an evil person anymore. The panopticon feeling worn off after a couple of experiments with both participants: they got accustomed to being watched. (Appendix 2: Dream recollections)

Figure 15: Somatosensory stimulation with the vibrations of a pedometer

Next step

To conclude, vibrations applied directly in REM seem to have the same level of impact as pre-stimuli when it comes to influencing dreams. As a result, it should be taken into account when comes the time of the test of the interaction concept. A “vibration-suit” kind of device (that apply pressure on any part of the body, without the knowledge of the dreamer) might be a way of applying a satisfying stimuli without fear of waking the person up. Such a suit has already been crafted on a chicken, by Adrian Cheok in 2012. A chicken was equipped with a jacket embedded with vibrating elements. Connected through the internet to a chicken doll, the jacket vibrated every time the owner would pet the doll (Figure 16). The person will know something will happen wearing the suit, but not pinpoint in advance where it will happen, letting space for anticipation and surprise. Furthermore, vibrations can be applied in different intensities, thus triggering different responses in the dream.

Figure 16: Adrian Cheok testing the link between the doll and his chicken jacket

Concept sketching: generating novel interactions

Along all the experimental phase, I produced concept sketches and design scenarios of possible interactions exploiting the experiments’ findings (see Gaver & Martin, 2000). Would the final concept use only pre-stimuli to influence dreams? What kind of dreams would be targeted (hallucinations at sleep onset, NREM, REM (Stickgold et al., 2001)? These concept sketches helped me explore the design space and discover what it had to offer. The main purpose of these concepts was not for them to be “good” or “desirable” design proposals. It was to explore the possibilities, play with the potential scenarios and produce a decent quantity in order to establish a “perimeter”. These sketches can be used to start a dialogue or can be used as inspirational material on what a playful interaction using sleep can look like. (Appendix 3: concept sketches).

The chosen concept is the “Voodoo Sleep Pet” (Figure 17). In this concept, the physical object is divided into 2 parts: one is “the stimuli pajamas”, worn by the dreamer. The second part is “the ball”, held by the waker. It is a tangible representation of the sleep of the dreamer, and emits sounds and visual feedback based on the emotional states and sleep stages of said dreamer. The ball tells the waker when the dreamer is dreaming, and in turn the waker interacts with the ball, which transmits the feedback directly to the stimuli pajamas. The pajamas vibrate accordingly, influencing the dreams of the dreamer. Due to the remaining time for the thesis and my programming skills regarding such a complex prototype, I decided to use some variation of the wizard of Oz technique to test it (Dow et al., 2005). It allowed me to play a game of “Let’s pretend” with both testers, leaving the concept open to changes and suggestions from both testers. The user-testing session turned into a co-creation kind of session. It also allowed me to gather intel from the waker’s point of view, after so many experiments focusing only on ways to interact with the dreamer.