- 1 -

BioGraphr: Science Games on a Biotic Computer

Figure 1. We developed the BioGraphr which uses millions of single-celled organisms (Euglena gracilis) in its core. These living cells form collective bioconvection patterns when stimulated with light patterns. Human users experience this emergent multi-cell behavior, thereby bridging phenomena at very different length-scales. This biotic computer enables game play, science, and art.

ABSTRACT The advancement of biotechnology enables novel types of interaction devices and alternative computers. How to design effective human interactions on these technologies is still largely unexplored. Here we present the “BioGraphr,” an interactive tablet device that enables playful experience and interaction with millions of microorganisms at a millimeter scale: Light patterns (images) are projected into a mini-aquarium containing phototactic Euglena cells, which then arrange into complex dynamic bioconvection patterns within seconds. We characterize the bio-computational properties of the BioGraphr; and we design pattern matching games and other interactive scientific and artistic activities. Our user studies demonstrate fun and meaningful gameplay, in particular how learning about microscopic biology can be naturally coupled to its “computational” substrate, making the case for “real” rather than “virtual” reality. This “BioGraphr” is low cost and easy to reproduce for museums and even children, and we derive general HCI design lessons for games on biological machines.

Author Keywords Human-Biology Interaction; Interactive Biotechnology; Tangible Microscopy; Biotic Game; Biotic Processing Unit; Microscopy; Alternative Computation; Education; Tablet

ACM Classification Keywords H.5.2. User Interfaces; K.3.m Computers and Education

INTRODUCTION, MOTIVATION, AND GOALS We currently witness a biotechnology revolution, which enable novel human interaction devices and alternative computing architectures that are either inspired by biological processes or directly utilize them [5,21,22,27,31,46,47,49]. Recently, several such “human biology interaction” (HBI) devices [24] have been developed and successfully tested. Examples include biology cloud labs [17], biotic games [39], museum exhibits [24,25], and citizen science [23], all of which merge biology, computers, and users on various levels. All systems incorporate playful approaches as games are regarded to have high educational potential, and are also good use cases for HCI research in general [6,8,11,15,48]. Previously, these systems focused on individual cells or small assemblies (or even molecules), while the interaction with millions of cells and associated phenomena at the millimeter scale has not been explored.

Therefore we designed the “BioGraphr,” a “biotic computer” that utilizes the collective behavior of over one million Euglena gracilis (Euglena) cells (Fig. 1): The phototactic behavior of these microorganisms [14,29,43] generates striking patterns when illuminated with homogenous or structured images of light. These patterns are a form of bioconvection, i.e. “The motion of large numbers of small organisms in a fluid.” [43] This bioconvection inside the BioGraphr constitutes a non-trivial computational transformation that bears analogies to edge detection and photographic processing (Fig. 1). In contrast to conventional electronic computers, here the information processing is carried out by millions of living cells that actively move in 3D space due the external stimuli, and that also interact with each other via shading and hydrodynamic effects at the local and global scales [43]. We turn this biophysical process into a practical and robust human-

Paste the appropriate copyright/license statement here. ACM now supports three different publication options: • ACM copyright: ACM holds the copyright on the work. This is the

historical approach. • License: The author(s) retain copyright, but ACM receives an

exclusive publication license. • Open Access: The author(s) wish to pay for the work to be open

access. The additional fee must be paid to ACM. This text field is large enough to hold the appropriate release statement assuming it is single-spaced in Times New Roman 8-point font. Please do not change or modify the size of this text box. Each submission will be assigned a DOI string to be included here.

- 2 -

interaction device via a touch-screen tablet with integrated camera and external projector to image and stimulate the cells housed in a “mini aquarium.” Hence, a closed feedback loop is established between the human user and the microorganisms.

To test the affordances of the BioGraphr and to elucidate general principles for human-biology interaction design, we designed and play tested games for fun, but also with the aim to foster informal science education (as a side focus we also explored applications for scientific inquiry and artistic expression). Games motivate players to explore and learn the “laws” of the “world” the game is set in [20], and we therefore rationalize that games on the BioGraphr should lead to a direct educational benefit regarding biology, such as via direct knowledge acquisition, or preparation for future learning [10,20,35]. This is in line with results from augmented reality systems [7,13] that expand the educational potential of digital technologies [2,26] with benefits like increased motivation and engagement [32,36]. We are also motivated by the persistent need for improved formal and informal science education [1,33,37].

We follow many of the design lessons for interactive biotechnology that have been proposed previously [24]. At the same time, we aim to make new contributions to open design questions in the field, such as whether meaningful games can be designed at the multi-cell scale; whether devices with living matter can have robust plug-and-play convenience like electronic devices – and at cost and complexity even suitable for children; or what game rules support educational objectives regarding the underlying biological phenomenon. When assessing a new game platform and genre, for example of whether it is fun or whether there is “meaningful play,” different schemes have been proposed [19,40,41], which we use as guide, but ultimately choosing the specific tools we believe best fit our present intentions.

This paper is structured along five goals: (1) To design a robust and practical device (The BioGraphr) that enables human interaction with the emergent bioconvection patterns of millions of cells; (2) to characterize this “biotic computation,” which then provides constraints and opportunities for interactive application; (3) to design meaningful games and other applications; (4) to assess the user responses to the system and its applications, in particular of whether the games are perceived as fun and provide meaningful play, and whether they have educational value in transferring knowledge or excitement about biology to the player; (5) to explore dissemination pathways of the setup to the public. We end with lessons learned and future perspectives.

1) DESIGN OF THE BIOGRAPHR (BIOTIC COMPUTER)

Bioware: Underlying biological principles We developed the BioGraphr which uses the single-celled photosynthetic Euglena gracilis (Carolina, #5867092) in its

core. Euglena is a microorganism that is commonly found in freshwater ponds (Fig. 2a). Euglena has been used widely in educational settings due to its robustness, safety [28], and relatively large size (~ 50 µm in length) which allows students to study them with conventional light microscopy. They are phototactic, i.e., orient and swim in response to light. This phototaxis is achieved by a photosensitive organelle (the “eyespot”) coupled to a motile flagellum at its anterior pole (Fig. 2a), which propels them forward. In nature, these cells need to retain optimal light conditions, i.e., sufficient light to perform photosynthesis but not so much to overheat or sustain UV damage. Hence, using positive and negative phototaxis at different light levels ensures they remain approximately 2-4 cm below the pond surface. Here we use relatively strong light to induce the cells swimming away from the light source. This photophobic response is virtually instantaneous, can be observed within seconds, and therefore allows for real-time interactions with human users [24]. This quick response results in a pattern-matching functionaliy that is the main mode for the games and activities described herein. (See Figure 2c stripe side, left)

Figure 2. The BioGraphr utilizes the collective swimming patterns emerging of millions of phototactic microorganisms in response to light. a) Single-celled Euglena. b) Schematic of emerging spatiotemporal Euglena patterns: Light projected from below stimulates cells to swim to darker regions and to the surface, cells also accumulate to shade each other, fluid with high-cell density may sink down leading to circulating bioconvection patterns. c) Experimental time course for stripe and homogenous light stimulus pattern.

In addition to the photophobic behavior that leads to the quick alignment to projected light pattern, a larger-scale emergent pattern (Fig. 2 white side, right) can evolve at sufficiently high cell densities and light intensities: If a strong light source is placed below the organisms, all cells swim upwards and agglomerate at the surface, where they form patches of very high density as they shade each other. As the cells are slightly denser than the surrounding water, the fluid as a whole now gets denser at the top. For

- 3 -

sufficiently high cell density the top fluid can even become unstable and starts sinking down at a higher speed than the Euglena cells can swim upwards. Thereby the denser fluid pulls the cells downwards, while at the same time less dense fluid, i.e., with a much lower Euglena density, is dragged upwards. This leads to vertical convection rolls that look phenomenological similar to thermal convection rolls [4], but in this case the energy is actually generated by the cells themselves. This is a more extreme example of bioconvection, which we will refer to as “circulating bioconvection” in this paper. These macroscopic patterns become visible within seconds after the strong light is turned on. This can be understood from the swimming speed of Euglena (about 0.1mm/sec [29]) and the depth (3 mm) of the used aquarium. Note, the overall light intensity is not harmful to the organisms; the cells in the mini aquarium remain responsive for weeks without significant accumulation of dead cells or decline in systems response.

Hardware and software In order to utilize these emergent multicellular patterns in game play, we implemented the BioGraphr (Fig. 3): The main functional components are an LCD DLP projector (IncrediSonic Pico Mini) that projects light onto the aquarium containing Euglena (Fig. 3a) and an unmodified Samsung tablet (Samsung Galaxy Tab Pro 12.2, Android 4.4 Kit Kat OS) for image acquisition (via the back-facing camera) and user interaction. The aquarium holding Euglena (Figs. 2a, 3a) was prepared from two acrylic discs and one acrylic ring with a diameter and height of 3 cm and 3 mm, respectively. A small hole was drilled into the side of the ring before acrylic glue was used to stick the pieces together. A suspension of Euglena containing approximately one million of cells was then injected through the hole. No sealing was applied. Such an aquarium remains functional (and the Euglena alive) for ~4 weeks if maintained at a natural day/night rhythm. For homogeneous specimen illumination the projector is placed directly under the Euglena (Fig. 3d,e), and we use a physical servo-shutter that controls whether light from the projector reaches the Euglena or not. This shutter is controlled by an application on the tablet communicating with an Arduino UNO via Bluetooth (all code in SI). The tablet also communicates wirelessly over Google Chromecast with the projector. The projected light patterns have a resolution of <100 µm at the site of Euglena. Conceptually, this linear forward loop (projector, biological material, camera) represents a Biotic Processing Unit (BPU) [17,39]. Hence a closed feedback between human user and biological material can be established, i.e., the user can stimulate the cells with light while seeing them at the same time.

The BioGraphr operates as follows: The projector shines a light pattern on the microaquarium. Just before the app takes a picture (typically every 4 seconds) a signal to move the servo (i.e., to close the shutter and block the light pattern) is sent over Bluetooth to the Arduino. After the image is taken, the shutter opens again and lets the light

pattern reach the aquarium (Fig. 3d). Note, instead of using the physical shutter, the projector can also be set periodically to dark when the image is acquired. We decided to use the physical shutter as it led to better images with a more homogeneous background and minimal glare. The graphical user interface and the hardware control for imaging and projection are written in Android Studio. The aquarium sits on a slider that enables removal of the aquarium. This is important, as before starting a new game, the cells need to be “reset“ (i.e., gently shaken to redistribute the Euglena and thereby erase the previous pattern), at the same time it enables the user to experience the biological matter more physically.

Figure 3. The BioGraphr consists of a projector, a tablet with integrated camera, and an aquarium with Euglena cells in between: a) The aquarium containing approximately one million of Euglena cells; b,c) placing the aquarium under a microscope reveals individual Euglena, as well as bioconvection patterns in the case of strong bottom microscope illumination. d,e) The complete setup with all components is simple and easy to reproduce. Elements like shutter and observation ports via an optical beam splitter additionally augment the biological content for the user. f) The aquarium is inserted with a slider from the side. g,h) The Euglena (and the projected pattern) can be observed directly from the side or through the objective; note a hexagonal pattern of black dots is projected.

We also include a number of design features that allow users to better understand the inner workings of the whole setup and also to facilitate curiosity about biology. To observe the organisms by eye and with the external microscope, the aquarium can be taken out of the

- 4 -

BioGraphr (Fig. 3a). As the aquarium is reset by gently shaking, this removal process is a necessity and becomes a natural task for the user. The user may observe the aquarium outside the BioGraphr with a conventional microscope augmenting the system (Fig. 3c), or live in the system through the eyepiece and beam-splitter (Fig. 3h). Additional features such as transparent casing and integrated visual observation ports (Fig. 3f-h) allow observation of the BPU’s operation.

2) PERFORMANCE OF THE BIOTIC COMPUTER

We interpret and characterize this bioconvection pattern formation as a form of dynamic computation: A light input pattern is transformed into a Euglena output pattern over time (Figs. 2c, 4a-d), where the transformation is actually non-trivial in the sense that it is not just an identity transformation. This bioconvection computation can be loosely compared to the neuronal processing of images in the human visual pathway [34], where images projected onto the retina are transformed into complex spike patterns in different neuronal layers. Characterizing this computation is both important for understanding the affordances and limitations of any human interaction (game play) but also intriguing intellectually.

To characterize the response to spatial light stimuli, stripe patterns with different widths were projected. Small stripes (0.5 mm) led to clear domains where Euglena agglomerate in dark areas and the stimulus pattern is imprinted (Fig. 4a). For large stripes (3 mm), agglomeration happened initially both in dark areas as well as at the edges (Fig. 4b,c), and where the Euglena density inside the dark areas (which actually marks the center of the stripe) decreases over time (* in Fig 4b,c). Additionally, some smears were observed in bright areas, due to circulating bioconvection setting in. Hence the Euglena response initially detects edges and centers of stripes (given of sufficient size), then evolves into pure edge detector, and eventually vertically circulating bioconvection takes overs.

To characterize the dynamic evolution, homogenous white light was projected. After 5-10 seconds circulating bioconvection patterns started to form (Fig. 2c right side). These patterns fluctuate over time and remain similar for at least 10 minutes. After the light is switched to black (or the shutter is closed) the patterns disappear within 30-40 seconds. Figure 4a,d shows the temporal evolution with the patterning amplitudes increasing fast initially, and then becoming smaller again.

Regarding colors and light intensity, while a full white (100%, approx. 12,000 lux) projection pattern induced strong vertically circulating bioconvection patterns after only 30 seconds, it took 40 seconds at 80% brightness, while at 60%, only weak circulating bioconvection patterns were formed and only after 3+ minutes. Lower intensities did not result in any circulating bioconvection patterns at all. Similar effects were observed for different colors (Fig.

5a): Pure white and black result in the fastest, while green and red take progressively longer (approximately 10 sec, 15 sec, 40 sec, 60 sec, respectively). Hence color and intensity can be used to fine-tune the temporal patterning aspects.

Figure 4: Euglena bioconvection in response to light represents a non-trivial computational process, loosely analogous to visual image processing in the brain. a) Small stripes (0.5 mm) projected on initially homogenous Euglena solution shows evolution of a “photograph” over time, where bright light regions lead to low Euglena density (and vice versa); blue lines are Euglena density averaged vertically over the image. b) In contrast, large stripes (3 mm), lead to pattern more like an edge detector, furthermore secondary peaks initially develop insight bright regions (*), that later disappear. c) Enlarged signal from large stripes. d) Amplitude evolution of the pattern in a, showing fast emergence of pattern, and later decay.

Figure 5: Euglena response after 30 seconds of patterned light stimulus each. a) 4-in-1 patterns demonstrate responses due to light intensity and spatial structures. Complex patterns and fine tiling can be “photographed.” b) 4-in-1 patterns demonstrates patterning responses due to white and blue light, but not to red or black. c) Superposition of two patterns projected alternately at 1 Hz leading to a near linear superposition.

- 5 -

Finally, we assessed the effect of projecting multiple patterns serially, and the effect of switching frequency on bioconvection pattern formation. We alternately projected a large stripe pattern (3 mm) and its exact color-inverted pattern at various frequencies (not shown). For switching at 10 Hz, no matching stripes were formed by the Euglena, but vivid circulating bioconvection patterns emerged after 30 seconds as would be for pure white light. At 2 Hz, Euglena underwent simple bioconvection and agglomerated in thin stripes (1 mm) at the borders of the projected stripes. Similar observations were made 0.2 Hz, while for switching times above 2 minutes, the Euglena had enough time to move and agglomerate in the dark areas, thus matching the projected patterns as in Figure 4b (but also exhibited some circulating bioconvection). Furthermore, alternating between distinct patterns (Fig. 5c), we obtained a direct superposition or more complex patterns, depending on the switching rate. If the last pattern is projected for longer times, this pattern will eventually dominate (with interference from circulating bioconvection). In conclusion, the resulting Euglena patterns are history dependent and usually not just simply linearly superimposed; this history dependency also implies that patterns are not commutative in general.

Comparing pattern formation multiple times on the same sample (after resetting) at different days or with other, equivalently prepared samples can reveal quantitative variations in the response, but the overall qualitative behavior is very robust, certainly enough for the present game and science applications. These “noisy features” are normal for analog and alternative computing paradigms [30] and highlight the biology to the user.

In conclusion this biotic computer transforms light patterns into emergent Euglena pattern in a reproducible manner as one would expect from a computational device. These responses are time, space, and path dependent, and we assessed optimal regimes for meaningful human interactions: patterns in the millimeter range are replicated by the microorganisms within seconds to minutes, circulating bioconvection patterns are formed in large bright areas, and the order of projected patterns affects the output. With these bio-computational characteristics at hand, we can do rational game design, especially tuning game challenges; and awareness of these characteristics leads to play strategies in one- or two-player games.

3) INTERACTIVE HUMAN-BIOLOGY APPLICATIONS

Activity overview menu We developed multiple interactive human-biology experiences (Figs. 6,7) for the BioGraphr. On the starting menu the user sees a live view of the Euglena, which are magnified through the camera about 4x, hence individual Euglena cells are not resolved on the screen. Users can choose from the main activities of Exploration, Games, and Science activities (Fig. 6a). The game category has a submenu (not shown) allowing selection of one vs. two player games, each offering multiple levels of difficulty. We explored various game ideas on this platform and eventually found that a “pattern guessing game” lends itself to single and multiplayer games that optimally utilize the features of the biotic computer. These types of games also make it easy to create game levels of different difficulty and without the need for complex automated image analysis.

Exploration and artistic expression The exploration tab (Fig. 6d) is intended as an open

Figure 6: Screenshots of the various user interfaces: a) Home screen. b) 2-player game (1-player game interface is shown in Fig. 1). c) Science tab. d) Explore tab.

Details on the Explore tab: The user sees the Euglena pattern live in the center. Colorful patterns drawn by hand over the live view (left) or preset patterns (right) can be selected by the user, projected onto the Euglena, and the resulting pattern observed, demonstrating response to white and blue but not to red. (Note: The drawing was actively misaligned and scaled to illustrate underlying pattern formation.) This interface enables free exploration, artistic expression, and was rapid prototyping of games and other applications. Additional buttons enable manual image focusing, opening and closing of the shutter, toggling the drawings, and the ability to save and share images.

- 6 -

interactivity to freely explore the system and biological responses to patterns and colors. It shows a live view of the Euglena pattern and the user can select pre-made patterns or colors to draw free-hand expressive patterns on top of the displayed Euglena. The user may use the automated image acquisition mode, where the shutter is closed every 4 seconds to acquire an image, or alternatively use a manual shutter mode that also allows for live interaction and observation of the Euglena. In live mode the projection and the shutter may be toggled manually. To enhance the view of the Euglena, the drawn images can be toggled on/off on the image (while always being projected onto the Euglena). This open interface allows the user to familiarize herself with the spatial and temporal response to light in a “live mode,” rather than still images (Fig. 6). Different icons and instructions guide the user; we aimed for simple and intuitive controls.

One-player games We intended to develop a single-player game that would last about a minute and could be easily adjusted in difficulty. In our final game design (Figs. 1,7), the computer chooses one pattern that is projected onto the Euglena, and the player is presented with a selection of four possible patterns. The player now needs to guess from this selection the pattern that is actually projected. Selecting the correct pattern advances the player to the next round, where a new pattern is randomly selected and projected; selecting the wrong pattern results in a time penalty of 10 seconds and the player still has to keep guessing. The game ends after three rounds. The game score is based on total time (including a time penalty for wrong guesses). To achieve the best score, a strategy combining speed, risk, and experience is required.

Figure 7: In the one-player game the player observes the evolution of Euglena patterns, from which she has to guess which of four randomly generated image patterns is projected. Typical sequence over 3 rounds is shown. Note, as Euglena are not “reset” between rounds, patterns blend into each other. The reader is invited to guess the correct answers.

Two-player games We intended to design a fast paced two-player game on a minute time scale, and where the players actions directly affect the stimulus to the Euglena and consequently the resulting bioconvection patterns. We derived at a game (Figs. 6b, 8) where each player is presented with four random patterns to choose from. Both players can see all eight patterns, which might affect their own selection. To choose, each player selects one of his four patterns in secrecy (by blocking the opponents view by hand). Then players press start, which swaps the four displayed patterns, i.e., each player can now choose from one of the four choices the opponent had. At the same time, both patterns are projected simultaneously onto the Euglena, i.e., the resulting bioconvection pattern is a more complex superposition due to the effects of two patterns. Every 4 seconds the shutter is closed to acquire an image that is then shown to the player. The players now have to guess which pattern the opponent has selected and select the matching pattern. If guessed correctly, the player gets 1 point and the round ends. If guessed incorrectly, the player loses 1 point, the incorrectly guessed pattern is x-ed out, and the round continues until one player guesses correctly. Multiple rounds are played until one player reaches 5 points to win the match. This game leads then also to the psychological game of outguessing one another, i.e., “You think, that I will choose this pattern because it is hard, therefore I will do something else…”. Note, that in the two-player game the players actively select the projected patterns and thereby directly affect Euglena behavior.

Figure 8: In the two-player game each player secretly chooses the pattern (out of four randomly generated ones) that will be projected. Both patterns are superimposed leading to the evolution of complex Euglena pattern. The player who first guesses the opponents pattern correctly scores one point; wrong guesses are penalized. More subtle “psychological” strategies start to emerge among players. Note: Euglena are “reset” between rounds, hence only patterns from same round blend into each other. The reader is again invited to guess the correct answer for both players.

- 7 -

Science activities In order to easily demonstrate the general biological response to patterns and colors, we integrated a simple and directed science tab (Fig. 6c): two experiments with four predefined patterns within a single image each allow side-by-side comparison of the effect of projecting either different colors or different patterns. Additionally, more explanations and scientific facts about the system and Euglena are revealed via the “Learn more” button. Figure 6c displays a screenshot of the color experiment with the Euglena having responded accordingly to white and blue but none to red or black; Figure 5a,b shows the results to both experiments.

4) HUMAN-BIOLOGY INTERACTION USER STUDIES

User study goals, design, and logistics The primary goals of the following user studies are to: (i) understand general affordances of the setup and how users respond to it; (ii) determine whether the activities, especially the games, are perceived as “fun” and interesting, and whether meaningful game play emerges; (iii) whether the games have a learning curve; (iv) how game difficulty can be adjusted, and (v) whether game play leads to inferences and learning about the underlying biology. From this we aim to identify general design principles for human-biology interactions in general.

The logistics of the user studies were as follows: (i) We conducted multiple pre-studies in the lab during the iterative design process. (ii) Over multiple days we set up our system at various sites on the University campus. The setup was complete with signs, information posters, and a microscope. Given the location, most of these users were students and postdocs. In most cases, users (single or in pairs) were introduced to the system, assessed with a pre and post-activity quiz about their knowledge of Euglena, phototaxis, and circulating bioconvection, and asked whether they have played a game involving microorganisms before or not. They then played one or more games; some explored the other activities (Fig. 6) as well. We also asked for verbal feedback and took notes on their comments. Single- and two- player games were tested. (iii) We went to a local science museum, drawing a diverse population, including children and adults. Typically, one of us was mediating the setup, and the other conducting interviews and taking notes. In total, over 60 people engaged with the system. Typical interaction time was between 1 and 5 minutes (some people who used the explore tab stayed 10+ minutes), while playing a single game takes about 1 minute. Note: The results in following sections are sorted based by research question rather than individual study.

i) Public responses and overall acceptance All users expressed no difficulty switching between tabs and applications (Fig. 7). While a tablet as a gaming device is well known by all users, the black box’s appearance (Fig. 3d,e), including the Euglena, projector, and the moving and

noisy shutter, were new to them. Generally, we facilitated the first game round and reset of the aquarium. After that, players were able to play and reset independently.

From the recorded verbal comments and the conducted post interviews, a few common themes emerged (Table 1): Nobody had seen something similar or played biotic games before (“That is special.”). Many users expressed positive surprise (“I can see them move!”) and fascination toward this gaming and learning concept using living microorganisms and felt it was novel. Still, while being immersed in the games, it appeared that the users quickly forget the underlying biological basis, and only reflect on it in between and after the games. Hence the setup and its application capture people’s attention and are easy and convenient to use.

Category Typical remarks Concept and Technology: BPU, projector, optics, Euglena.

“That is special.” “What is that noise?” “What does the Arduino do?” “Are there any lenses?”

Scientific Observations: Biology, dynamics of responses, effect of patterns and light. Desire to understand.

“It forms stripes!” “I can see them move!” “They move pretty fast!” “Blue is similar to white.” “Do colors matter much?” “It changes now.” “What is it really?”

Computer-Human Interaction: Wish for implementation.

“You can control them.” “What else could you do?” “Can it solve problems?”

Games: Response time, interactions between players, game strategies, game bystander’s comments.

“I win!” / “I beat you!” “It’s so fast” “Stripes are easy.” “The first pattern is easier.” “Try this pattern” “I know it!” “Competition is exciting.”

Expansion: Expand in free play. Interest in BPU. Wants to replicate the system.

“Want to do my own experiments.” “Where do you get Euglena?” “How much does it cost?”

Table 1. Categories of comments made by participants.

ii) Emergence of game strategies and meaningful play The single-player game (Figs. 1, 7) was played in a straightforward manner. Overall rating of the question “How did you like the overall activity?” on a five-point Likert scale was above 4.7 (σ = 0.5; n = 10). While playing, many users discovered that pattern guessing gets harder the longer one waited (“That was very hard.”) and therefore came up with the strategy of guessing sooner and risking a penalty rather than waiting too long. Taking too much time will result in a “burnt-in” pattern and thus make the subsequent pattern more difficult to guess. The best scores (27 seconds) were achieved when a player waited 1-3

- 8 -

frames (i.e. 4-12 seconds) for each pattern. A typical full game with three guesses lasts < 60 seconds. It is interesting to note that many players seem really immersed in the game while playing, thereby forgetting the “world around them,” which also included forgetting about the underlying biology. However they reflect afterwards on the general implications of what they just experienced.

The two-player game (Figs. 6b, 8) received very positive feedback in all study settings and players clearly enjoyed playing it (“I beat you!”). We also ran an elimination tournament with eight players, i.e., 7 matches total, each consisting of typically eight rounds, i.e., until one player got 5 points. Players became immersed easily into the game, and competitive behavior was observed. To develop a wining strategy, one has to realize that some patterns are easier to recognize than others. Many players indicated that stripe patterns are easier to guess than dot patterns, and small patterns are easier to guess than large patterns and therefore players often selected large dot patterns for their opponents to choose. Upon noticing this fact and through mutual discussion between players, secondary psychological strategies emerged, as commonly observed in game theory [3]: Players selected a pattern similar to what they expected their opponent to choose, so that the resulting pattern was harder to recognize. For example, one player stated that she would never choose the stripe pattern if presented together with three dot patterns. Since her opponent knew this, he was sometimes able to guess correctly without even looking at the Euglena’s response. Experienced players could guess the opponents choices even before the pattern had become visible. The development of game strategies implies significant engagement in the single and two-player games, and we consider both single- and two-player games as successful games.

iii) Repeated game play improves playing skills Based on these observations and verbal feedback, players seemed to become more skilled during the game by recognizing specific features of the game mechanics, including those that are due to underlying biological dynamics. We set out the following test to show in-game learning by demonstrating improvement of scores quantitatively: We asked 10 users to play 5 single-player games each (Fig. 7). Each game consisted of 3 projected patterns with 4 patterns displayed to guess from. The first and last game were loaded with fixed patterns and remained the same for all test users, all other patterns were generated randomly based on the following rules. Game 2 was a sequence of 3 pattern games going from large, to medium, to small features. Game 3 went from small, to medium, to large. The fourth game was completely random and we therefore expected it to be the toughest. Figure 9 summarizes times, penalties (i.e. mistakes), and scores. We found a significant improvement (p < 0.001) from game 1 to game 5 which are identical; (testing confirmed, that players did not realize that the last game was exactly the

same as the first) as measured both in terms of time (38.8 to 30.3 seconds) and the number of errors (9 to 2 total), while no significant game-to-game learning difference was observed for all the other games. The two conclusions from this analysis are that players acquire in-game skill while playing this game, and that the game is neither trivially simple nor arbitrarily difficult.

Figure 9: Learning curve for repeated play (learning = lower game score). a) All players achieved better scores in game 5 then in game 1. From internal testing we expected the difficulties to increase as follows: 1 = 5 < 2 < 3 < 4. This is a desired effect of deliberate game design as it continuously challenges the player. The 5th game acts as an assessment control. b) Comparing first and last game shows that all players (10/10) improved their game scores.

iv) Biological properties allow tuning of game challenge Next, we wanted to characterize a few “biological knobs” that affect game difficulty, hence enable more rational game design: (i) In line with comments made verbally and on the feedback forms, the first pattern within each game is easier to guess, i.e., players guess it faster and make less mistakes: Pooling all data (same as Fig. 9) for the 1st, 2nd, and 3rd patterns over all games lead to the following results: the 1st patterns were guessed significantly faster than 2nd and 3rd patterns (12.3 ± 0.14 s, 13.4 ± 0.53 s, 14.8 ± 0.38 s; p < 0.05). Also significantly fewer mistakes were made in the 1st pattern compared to 2nd or 3rd (7, 26, 19; p < 0.05). We attribute this to two facts: First, the Euglena react faster from a perfectly dispersed state. Second, the superposition of later patterns on a previous one makes it harder to recognize the new one. Therefore, the average guessing times increase within a game from pattern 1 to 2 and from pattern 2 to 3. This is also in line with the biophysical response time of pattern development (Fig. 4d). (ii) Looking at the itemized times for small, medium, and large patterns for one player games, there was a trend that it took players longer to recognize medium patterns (14.4 ± 4.7 s) than large (12.0 ± 4.1 s) and small (11.9 ± 4.2 s) patterns (difference is not significant). Interestingly, no user indicated that medium patterns were tricky – while some stated large, and others stated small to be hard. All users agreed however that stripe patterns were easier than dot patterns (no timed data were tracked). (iii) The different slowed down response times due to “colder colors” (red, blue; Fig. 5a) or lower intensity increases user response time. Interestingly, very experienced players can recognize

- 9 -

the color based on the absence of emerging pattern. (iv) Additional adjustment in challenge is achievable by changing the intensity of the light or the shutter speed, i.e., acquisition speed. With increased frequency of image acquisition, the players 1 also sees images that are more frequently updated, which then forces the player to pay closer attention to subtle changes in the Euglena images. Using these observations, the biophysical dynamics underlying the game can be actively exploited to tune game challenge.

v) Game play can lead to biology knowledge transfer Finally, we found significant evidence that people learn about biology and the technology of the setup while playing: Multiple users attempted to understand the overall setup as evidenced by their questions and comments (Table 1), and by looking at the setup. Also, everyone understood that these were living organisms and that this was not a simulation. Although most seemed to understand what phototaxis is, the distinction between pattern matching and vertically circulating bioconvection patterns was difficult to grasp for many users. No user had heard about bioconvection before; still, many expressed fascination after being provided with the biological explanations. The view through a microscope eyepiece helped to increase the understanding about how tiny the cells are and how fast they swim. Also, it was possible to demonstrate phototaxis in single drops containing Euglena. In conclusion, people can learn the most basic aspects (i.e., Euglena move away from light), but the bioconvection concept is much harder to grasp.

To further investigate of how people could learn about biology during game play, we setup the implicit learning goal “The Euglena eye spot is less sensitive to red than to blue light.” (see Fig. 5b for color response) Novice players were initially told that “Euglena respond to light” while explaining the setup – but no explicit mentioning of colors was made. Over 3 games (9 rounds total) of the single-player games (Fig. 7), the player was confronted initially with white patterns only, then successively colored pattern were mixed in. Here red colored patterns took much longer to develop, a fact that players noticed easily (“So they don’t react to red”), but some wondered whether the system is actually still working. In the last round all patterns were the same except for their color, i.e., the player had to distinguish based on response time of the color alone. Afterwards, we asked players the above question regarding color sensitivity of the eye – only a minority of players (2/8) succeeded in this task, but the majority did not. A better game design and more play experience would likely increase the success rate. In contrast, prompting novice users to the science and exploration activities (Fig. 6) appeared to be a more “time effective” way to make users recognize the color differences. Hence although these results demonstrate the potential for coupling educational content to game play on the BioGraphr, it still raises a

number of game design questions for the future of how to achieve this coupling much more effectively.

5) DISSEMINATION TO MUSEUM AND CHILDREN Presenting the BioGraphr in a local science museum as a guided game exhibit (Fig. 10a) demonstrated interest among wide age groups from 5 to 50 year olds; families with children as well as groups of school children interacted with the system, played the games (“Oh that is a fun game!”, “This is really hard”), showed curiosity (“So why do they respond to light?”), and where fascinated (“This is the best Arduino project I have ever seen.”). Hence we consider this a success, but we also noticed that it took 30 seconds and more for visitors to grasp that they were interacting with living cells, and that actual patterns of light were projected. This is in contrast to the previous TrapIt! Setup [24], which also utilized Euglena, but enabled the visitor to see and interact with individual directly in a live mode by drawing light over it on a touch screen, and which therefore drew visitors understanding much faster, and where time is at the essence for a successful museum exhibit [9,16]. This was further emphasized since children looking through the eyepiece get excited very rapidly (“There are all theses little worms!”); a response of that intensity we did not get when they looked at the tablet. Hence if the BioGraphr is to be developed into a museum exhibit, we suggest to introduce the visitor with the free exploration activity (Fig. 6d) rather than a game, and more importantly show a live view rather than images that are updated every 4 seconds; these are straight forward changes to make. We emphasize again that the BioGraphr from a pure game play perspective was very successful.

Figure 10: a) Visitors during exhibit in science museum interacting with regular microscope (foreground) and BioGraphr (background). b) Easy replication of system using Lego construction. c) Placing the Euglena chamber as an “aquarium” next to Lego minifigures conveys the concept of “very small organisms” to children. In order to enable wide dissemination, such as to schools and children’s rooms, we stripped the setup to its bare components, i.e., projector, aquarium, and tablet, and built the structural support from Lego bricks to hold these components in place (Fig. 10b). Here, the shutter function is executed only digitally, i.e., projector turning black. The culturing chamber can be laser cut, but even a conventional 35 mm diameter petri dish does work. Also, placing a minifigure with a magnifying glass (Fig. 10c) can convey the concept of “very small organisms” to children: When children in the museum discovered this minifigure they were particular excited (“Oh Lego - that is so cool - I want

- 10 -

to build something!”). Hence additional educational synergy could be realized when children build these interactive devices themselves, e.g., with Lego, as demonstrated previously for other media [38].

DISCUSSION AND FUTURE WORK Based on this work, we add the following recommendations and insights to previously established design lessons for Human-Biology Interaction (HBI) [24]: (1) Utilizing a biological phenomenon that relies on millions of cells rather than a few individual ones [24] significantly increases the robustness of the activity and also lowers the technological requirements, i.e., bioconvection is already observable without a microscope; and the larger, distributed pattern still form in a recognizable manner despite air bubbles or other contaminations inside the aquarium. (Actually, the air bubble is even a desired feature as it facilitates mixing of the culture.) (2) Our game design avoided the need for any computerized image processing to assess the state of the bioconvection pattern, which can be computationally expensive to implement robustly. Instead, the game questions always revolved around the image that was actually projected, which is easily “known” by the computer to allow for game scoring. Technically, all the image processing was off-loaded to the participating humans and Euglena. (3) A shutter that updates images every few seconds (rather than a live view) makes it much harder for users to realize that this is a real biological system. We do not know how fast an image update ideally would have to be, but we expect multiple times per second. We also expect that seeing the evolution of an abstract pattern without being able see the individual cells therein makes it again harder to recognize the living nature of the medium. (4) Placing human miniature figures next to the mini aquarium seems to help children to grasp both that there are living organisms and that they are small (“This is a fish tank for Lego!”). How much this scaling analogy really helps a child, and within what age range needs to be explored further. (5) Game and level design can facilitate learning about biology: Rather than telling people explicitly that single cells respond less strongly to red light, some players can infer this fact from their in-game experience regarding bioconvection pattern response times due to red stimulus patterns.

Figure 11: Extended MDA model for human-biology interaction, suggesting that successful game design in the digital space focuses the player’s experience on desired aspects of the underlying “biological mechanics”.

We can formalize this process of implicit learning while playing with a biological substrate (Fig. 11): Considering the Mechanics, Dynamics, and Aesthetics model (MDA) [18], we note that for both game designer and player two levels for biotic games exist: the digital and the biological. From a design perspective, both levels can be sculptured, but given the current state of technology this is much easier done on the digital than on the biological level. The opportunity (and challenge) for the game designer is now to develop a combined set of digital and biological game rules that lead to a “natural lens” for the player to experience the biological content.

On a more philosophical note, some readers may still question: Is this really a computer – or a just a biophysical process? We would argue that the same question could be asked about the visual pathway in the human [34] or tissue morphogenesis during embryonic development [45]. Especially for the HCI community, we believe it is worthwhile to frequently ask “what is computation” [12] – realizing that for any algorithm to work it has to be implemented on a dynamical physical substrate, whereby certain computer architectures are better suited for certain types of computation and applications than others.

We envision the use of this and other biotic computers in schools as do-it-yourself kits, and in public spaces like museums, with similar significant potential for formal and informal education as their electronic counterparts [1,42]. These activities also fit well with the formal learning content, such as the next generation science standards in the US, which for example emphasizes biology and scientific investigations in middle schools [44]. The Euglena inside the aquarium are safe to use for children [28], and robust and responsive for weeks, which makes the BioGraphr a practical device rather than an academic exercise. Code and building instructions are open source for replication and further development.

CONCLUSION We presented the “BioGraphr,” a new human-biology interaction device that utilizes the collective bioconvection patterns of millions of cells, thereby enabling immersive game play and other activities at the multi cell scale. This establishes a new game paradigm. The BioGraphr has proven to have a “plug-and-play” robustness approaching that of electronics (despite relying on living cells). The pattern matching games were successful as entertaining activities. Lots of creative design potential is now available to make other games (also to include more complex image analysis of the bioconvection patterns) and to unearth the full educational potential about the underlying biology.

ACKNOWLEDGMENTS The authors are grateful to our lab members for participating in initial user tests and providing valuable comments. We thank XXX museum for arranging the user test and providing useful resources. This research was supported by XXX.

- 11 -

REFERENCES 1. Bruce Alberts. 2010. Prioritizing Science

Education. Science 328, 5977: 405–405. http://doi.org/10.1126/science.1190788

2. Savilla Banister. 2010. Integrating the iPod Touch in K–12 Education: Visions and Vices. Computers in the Schools 27, 2: 121–131. http://doi.org/10.1080/07380561003801590

3. Tamer Basar and Geert Jan Olsder. 1999. Dynamic Noncooperative Game Theory. SIAM.

4. G.K. Batchelor. 2000. An Introduction to Fluid Dynamics. Cambridge University Press, Cambridge. http://doi.org/10.1017/CBO9780511800955

5. John I. Bell. 2003. The double helix in clinical practice. Nature 421, 6921: 414–416. http://doi.org/10.1038/nature01402

6. Zafer Bilda. 2011. Designing for audience engagement. Interacting: art.

7. Mark Billinghurst, Adrian Clark, and Gun Lee. 2015. A Survey of Augmented Reality. Foundations and Trends in Human-Computer Interaction 8, 2-3: 73–272. http://doi.org/10.1561/1100000049

8. Florian Block, Michael S. Horn, Brenda C. Phillips, Judy Diamond, E. Margaret Evans, and Chia Shen. 2012. The DeepTree Exhibit: Visualizing the Tree of Life to Facilitate Informal Learning. Visualization and Computer Graphics, IEEE Transactions on 18, 12: 2789–2798. http://doi.org/10.1109/TVCG.2012.272

9. Dorothy Lozowski Boisvert and Brenda Jochums Slez. 1995. The relationship between exhibit characteristics and learning‐associated behaviors in a science museum discovery space. Science Education 79, 5: 503–518. http://doi.org/10.1002/sce.3730790503

10. John D. Bransford and Daniel L. Schwartz. 1999. Rethinking Transfer: A Simple Proposal with Multiple Implications. Review of research in education 24: 61. http://doi.org/10.2307/1167267

11. Michael Brody, Arthur Bangert, and Justin Dillon. 2009. Assessing learning in informal science contexts. Commissioned Paper by the National Research Council for Science Learning in Informal Environments Committee.

12. Douglas E. Comer, David Gries, Michael C. Mulder, Allen Tucker, A Joe Turner, and Paul R. Denning, Peter J. Young. 1989. Computing as a discipline. Communications of the ACM 32, 1: 9–23. http://doi.org/10.1145/63238.63239

13. de Souza and Girlie C Delacruz. 2006. Hybrid Reality Games Reframed: Potential Uses in Educational Contexts. Games and Culture 1, 3: 231–251. http://doi.org/10.1177/1555412006290443

14. N.A. Hill and L.A. Plumpton. 2000. Control

Strategies for the Polarotactic Orientation of the Microorganism Euglena gracilis. Journal of theoretical biology 203, 4: 357–365. http://doi.org/10.1006/jtbi.2000.1090

15. Michael S. Horn, Erin Treacy Solovey, R Jordan Crouser, and Robert J. K. Jacob. 2009. Comparing the use of tangible and graphical programming languages for informal science education. ACM, New York, New York, USA. http://doi.org/10.1145/1518701.1518851

16. Michael Horn, Zeina Atrash Leong, Florian Block, Judy Diamond, E. Margaret Evans, Brenda Phillips, Chia Shen. 2012. Of BATs and APEs: an interactive tabletop game for natural history museums. ACM, New York, New York, USA. http://doi.org/10.1145/2207676.2208355

17. Zahid Hossain, Xiaofan Jin, Engin W. Bumbacher, Alice M. Chung, Stephen Koo, Jordan D. Shapiro, Cynthia Y. Truong, Sean Choi, Nathan D. Orloff, Paulo Blikstein, and Ingmar H. Riedel-Kruse. 2015. Interactive Cloud Experimentation for Biology: An Online Education Case Study. ACM, New York, New York, USA. http://doi.org/10.1145/2702123.2702354

18. Robin Hunicke, Marc LeBlanc, and Robert Zubek. 2004. MDA: A formal approach to game design and game research. Proc. AAAI Workshop on Challenges in Game, AAAI Press.

19. Ioanna Iacovides and Anna L. Cox. 2015. Moving Beyond Fun: Evaluating Serious Experience in Digital Games. ACM, New York, New York, USA.

http://doi.org/10.1145/2702123.2702204 20. Daniel Johnson, Lennart Nacke, and Peta Wyeth.

2015. All about that base: differing player experiences in video game genres and the unique case of MOBA games. ACM, New York, New York, USA.

http://doi.org/10.1145/2702123.2702447 21. Fanwei Kong, Liang Yuan, Yuan F. Zheng, and

Weidong Chen. 2012. Automatic Liquid Handling for Life Science A Critical Review of the Current State of the Art. Journal of Laboratory Automation 17, 3: 169–185. http://doi.org/10.1177/2211068211435302

22. Stacey Kuznetsov, Carrie Doonan, Nathan Wilson, Swarna Mohan, Scott E. Hudson, and Eric Paulos. 2015. DIYbio Things: Open Source Biology Tools as Platforms for Hybrid Knowledge Production and Scientific Participation. ACM, New York, New York, USA. http://doi.org/10.1145/2702123.2702235

23. Jeehyung Lee, Wipapat Kladwang, Minjae Lee, et al. 2014. RNA design rules from a massive open laboratory. Proceedings of the National Academy of Sciences of the United States of America 111, 6: 2122–2127. http://doi.org/10.1073/pnas.1313039111

- 12 -

24. Seung Ah Lee, Engin Bumbacher, Alice M. Chung, and Ingmar H. Riedel-Kruse. 2015. Trap it!: A Playful Human-Biology Interaction for a Museum Installation. ACM, New York, New York, USA. http://doi.org/10.1145/2702123.2702220

25. Seung Ah Lee, Alice M. Chung, Nate Cira, and Ingmar H. Riedel-Kruse. 2015. Tangible Interactive Microbiology for Informal Science Education. ACM, New York, New York, USA. http://doi.org/10.1145/2677199.2680561

26. Amanda Lenhart, Joseph Kahne, Ellen Middaugh, Alexandra Macgill, and Chris Evans. 2008. Teens, Video Games, and Civics. Washington, DC: Pew Internet; American Life Project.

27. Erkki Liikanen. 2003. Life sciences and biotechnology: a strategy for Europe. Communication from the Commission to the European Parliament. COM 27.

28. Robert A. Littleford. 1960. Culture of Protozoa in the Classroom. The American Biology Teacher 22, 9: 551–559. http://doi.org/10.2307/4439448

29. S. Machemer-Röhnisch, U. Nagel, and H. Machemer. 1999. A gravity-induced regulation of swimming speed in Euglena gracilis. Journal of Comparative Physiology A 185, 6: 517–527. http://doi.org/10.1007/s003590050412

30. Radhika Nagpal and Marco Mamei. 2004. Engineering Amorphous Computing Systems. In Methodologies and Software Engineering for Agent Systems. Springer US, Boston, 303–320. http://doi.org/10.1007/1-4020-8058-1_19

31. Rainer Pepperkok and Jan Ellenberg. 2006. High-throughput fluorescence microscopy for systems biology. Nature Reviews Molecular Cell Biology 7, 9: 690–696. http://doi.org/10.1038/nrm1979

32. Maja Pivec. 2007. Editorial: Play and learn: potentials of game-based learning. British Journal of Educational Technology 38, 3: 1–7.

33. Noah Podolefsky. 2012. Learning science through computer games and simulations. Studies in Science Education 48, 2: 237–240. http://doi.org/10.1080/03057267.2012.720770

34. Michael I. Posner and Steven E. Petersen. 1989. The Attention System of the Human Brain.

35. Chris Preist and Robert Jones. 2015. The Use of Games as Extrinisic Motivation in Education. ACM, New York, New York, USA. http://doi.org/10.1145/2702123.2702282

36. M. Prensky. 2005. Computer games and learning: Digital game-based learning. Handbook of computer game studies.

37. Linda Ramey-Gassert and Herbert J. Walberg. 1994. Reexamining connections: Museums as science learning environments. Science Education 78, 4: 345–363. http://doi.org/10.1002/sce.3730780403

38. Mitchel Resnick, Robbie Berg, and Michael

Eisenberg. 2009. Beyond Black Boxes: Bringing Transparency and Aesthetics Back to Scientific Investigation. The Journal of the Learning Sciences 9, 1: 7–30. http://doi.org/10.1207/s15327809jls0901_3

39. Ingmar H. Riedel-Kruse, Alice M. Chung, Burak Dura, Andrea L. Hamilton, and Byung C. Lee. 2011. Design, engineering and utility of biotic games. Lab on a Chip 11, 1: 14–22. http://doi.org/10.1039/C0LC00399A

40. Richard M. Ryan, C. Scott Rigby, and Andrew Przybylski. 2006. The Motivational Pull of Video Games: A Self-Determination Theory Approach. Motivation and emotion 30, 4: 344–360. http://doi.org/10.1007/s11031-006-9051-8

41. Jesse Schell. 2014. The Art of Game Design. CRC Press.

42. Heidi A. Schweingruber, Marilyn Fenichel, Board on Science Education, Center for Education, Division of Behavioral and Social Sciences and Education, National Research Council. 2010. Surrounded by Science: Learning Science in Informal Environments. National Academies Press.

43. Erika Shoji, Hiraku Nishimori, Akinori Awazu, Shunsuke Izumi, and Makoto Iima. 2014. Localized Bioconvection Patterns and Their Initial State Dependency in Euglena gracilis Suspensions in an Annular Container. Journal of the Physical Society of Japan 83, 4: 043001. http://doi.org/10.7566/JPSJ.83.043001

44. Elizabetz K. Stage, H. Asturias, T. Cheuk, and P.A. Daro. 2013. Opportunities and challenges in next generation standards. Science 340, 6130: 276–277.

http://doi.org/10.1126/science.1234011 45. W. Richard Stark. 2013. Amorphous computing:

examples, mathematics and theory. Natural computing 12, 3: 377–392. http://doi.org/10.1007/s11047-013-9370-0

46. S.J. Trietsch, T. Hankemeier, and H.J. van der Linden. 2011. Lab-on-a-chip technologies for massive parallel data generation in the life sciences: A review. Chemometrics and Intelligent Laboratory Systems 108, 1: 64–75. http://doi.org/10.1016/j.chemolab.2011.03.005

47. George M. Whitesides. 2006. The origins and the future of microfluidics. Nature 442, 7101: 368–373. http://doi.org/10.1038/nature05058

48. Nesra Yannier, Kenneth R. Koedinger, and Scott E. Hudson. 2015. Learning from Mixed-Reality Games: Is Shaking a Tablet as Effective as Physical Observation? ACM, New York, New York, USA. http://doi.org/10.1145/2702123.2702397

49. Lining Yao, Jifei Ou, Chin-Yi Cheng, Helene Steiner, Wen Wang, Guanyun Wang, Hiroshi Ishii. 2015. bioLogic: Natto Cells as Nanoactuators for Shape Changing Interfaces. ACM, New York, USA. http://doi.org/10.1145/2702123.2702611