Tools for Awareness – Awareness Lab design research blog

Driftwood tableware

Since the start of my Little Chris project I’ve used ChatGTP to help me with all kinds of aspects of machine learning. From which platforms to use, to analysing images, and setting up a dataset, etc. I’ve named the chatbot Rufus for easy conversation. For the next step in the Little Chris project, I want to build a driftwood-based dataset to train a LoRA focused on an Alexander-inspired property. The endless variate of shapes in driftwood keep amazing me.

Day 1

I’ve discussed with Rufus which properties are present in pieces of driftwood. We decided on the property gradients. Gradients refers to controlled transitions and gradual change [1]. Rufus gave advice on the variety and size of my driftwood collection in order to build a robust dataset such as: clear thick to thin gradient, subtle torsion (slight twisting), direction or sense of movement. These type of fragments will express the gradients property.

Day 2

Fig. 1. All the collected pieces of driftwood fresh from the beach

I went to beach to collect a variety of pieces of driftwood. I found about 50 interesting and beautiful pieces (Fig 1.).

Day 3

Fig 2. Part of the set of selected pieces, photographed from different angles.

I’ve asked Rufus how to photograph the wood for the optimal dataset. “The key is to photograph individual pieces against a neutral background, using soft diffuse light, and capturing each object from different angles […] Aim for around 30–40 distinct pieces, resulting in 80–120 images.” (View Fig. 2.) From this big set Rufus helped me to create a strong subset of 42 images with clear gradients properties. The focus was mainly on thick to thin pieces and subtle shape shifts.

Day 4

Fig. 3. Example of checkpoint outputs for the second model.

Training was harder this time then with the pebble dataset. That set was more extensive and uniform. I actually did two runs. One with the 42 images and captions, the second with a smaller set. Training a model is incremental. It yields different version of the model called checkpoints. You can test for the most useful checkpoint by generating an image from each checkpoint using the same prompt e.g.: woodGradient ceramic container, hand formed clay. WoodGradient being the class name of the model used during training (Fig. 3).

Fig. 4. Systematic testing of one checkpoint.

It was hard to determine the over or under training of the LoRA’s because this also depends very much on the prompt used. You can also play with the strength of the LoRA checkpoint. This determines the influence of the LoRA compared to the base model, in this case Stable diffusion XL Base 1.0. I tested three checkpoints systematically with the same seed and prompt but with different strengths (Fig. 4.). It seemed I had to increase the strength to 1.4 (1.0 being default). I decided to use a checkpoint from the first training.

Day 5

I decided to adapt the pebble prompt I used earlier. This useful prompt meant I could lower the LoRA strength to 1.0 or 1.1 for inference of the cups, bowls and dishes. Example prompt: woodGradient hand-formed ceramic bowl, open vessel with subtle gradient shape, wide mouth, shallow interior, driftwood-like material character, clay tones with natural colour transition, tactile hand-built surface, quiet asymmetry, grounded mass, soft daylight, neutral background.

I’m especially pleased with the bowls. The results are very diverse but all are beautiful and functional and they can be created in clay.

Reference

1. Salingaros, Nikos A.. “ch. 11 (19). Christopher Alexander’s 15 Fundamental Properties.” In Unified Architectural Theory: Form, Language, Complexity—a Companion to Christopher Alexander’s “The Phenomenon of Life : the Nature of Order, Book 1”, 125-130. Portland, Oregon and Kathmandu, Nepal: Sustasis Foundation and Vajra Books, 2013.

Making pebbles

Since a couple of months I’ve been wresting with and discovering the power of generative AI. My aim is to create custom models (LoRA’s) for image generation based on the fundamental properties formulated by design theorist and architect Christopher Alexander. The images serve as sketches for ceramic tableware such as cups, bowls and dishes. This tableware should support states of resonance in the user while eating lunch. (More on all these subjects later).

In this post I want to share some successful results from my Pocket Project* Making pebbles. In this project I create and use a dataset with pebbles and descriptions to create photo realistic images of pebble-like tableware. Pebbles adhere to the fundamental property of Simplicity and inner calm.

Fig 1. Part of my pebble collection and example for dataset image.

I have a large collection of pebbles collected over many years (Fig 1.). On day one I photographed all the pebbles in house, resulting in 63 images (Fig 2.).

The second day I asked ChatGTP to generate detailed captions (descriptions) of the pebbles. I also looked at automated generation but this was either to general or the focus was on the wrong elements. I need very detailed descriptions of the pebbles: their shape, structure, distinguishing features and colour. Only then can the model use these details in new images. Some of the captions were very poetic: A pebble with a compact, rounded triangular geometry, softly domed and grounded. The form feels stable and inward. Its surface is smooth and matte, warm brown in tone, marked by a lighter diagonal band that crosses the body, adding visual interest while remaining fully integrated into the overall coherent form.

Fig. 2. All pebble images of the dataset.

The third day I explored a new development environment. I was using ComfyUI at Runpod but I found the training and inference process lacking in transparency. Above all Runpod had kept me waiting for an available GPU for many hours, even in the mornings. I was fed up with it. I settled on vast.ai as a GPU cloud service. The pricing is good, they support the templates I need and the availability of the European servers is much better.

The fourth day I spend figuring out Kohya-ss, a set of scripts with a graphical user interface for training a custom model. It uses Stable Diffusion, a set of models for training and inference of images and text, as a base for working with your own dataset. The installation of the interface was easy, it was available as a template on vast.ai. But understanding all the settings and folder locations kept me busy for a day…

Fig. 3. Exploring the checkpoints, I chose checkpoint 8, third from the right.

On day five I could start the actual training with my dataset. In Kohya-ss it is easy to control every step of the training. I generated nine checkpoints, different stages in the training to assess the amount of training. Undertraining may result in sloppy, unrealistic images. Overtraining often generates artefacts. It is very helpful that you can let the program generate images during training to see how it is coming along. I chose the best checkpoint using a handy script which makes it easy to compare the checkpoint outputs (Fig 3.).

Fig. 4. Porcelain cup-like shapes from the same seed.

On day six I started the inference process, the actual generation of images. To do that I needed another set of scripts with a graphical user interface: Automatic1111. I installed this template on another instance on vast.ai. I asked ChatGTP to supply prompts, start settings and test protocol for generating images of the tableware items. This went surprisingly well (Fig 4.). A1111 creates images and provides an overview which contains the prompt, all the settings and the seed used. The seed is a random number attached to the specific image or set of images. It allows you to generate the exact same image. Generating interesting images of dishes was the hardest. The shape of a dish is very far removed from the morphology of a pebble. I needed to be much more explicit in the prompt. I ended up with a very useful set of images and a proof of concept of my new workflow (Fig. 5).

Fig. 5. Examples of a bowl, dish and cup generated with my own LoRA.

* A short one to seven day project with a clear and reachable aim, view the manifesto for more information.

Pocket Project Manifesto

Do you also think big and complex? I love connecting different domains, theories and practices. Alas this can turn bright ideas into top heavy projects that require many different partners, funding and a lot of time.
I realized I got stuck in complexity and making things became more and cumbersome. After an incubation period of a couple of months I’ve come up with the solution: Pocket Projects. Short projects with a clear goal that I can finish in a day or a week.
I’ve created a manifesto to celebrate my insights. And guess what? The manifesto was my first Pocket Project!
I hope this useful for you too. Have fun!

Interactive ceramics

Fig. 1 Test setup with conductive objects, Arduino Uno, capacitive sensor, piezo speaker and circuit.

Lunch with Resonant Dishes should be an engaging experience. Instead of a mindless, hurried routine this lunch should be worthwhile, interesting and playful. It should allow the user time to rediscover themselves and reconnect with themselves, the tableware and the food during every meal. To make this happen the lunch will be enhanced with rich tactile, visual and auditory stimuli. They will come from pottery itself and from digital or digitally enhanced sources.

Tactility

I want the pieces to respond to touch and to trigger certain events such as the playing of a sound. For ceramics to become part of such an interactive system they need to have conductive properties. I’ve experimented with lustre, a glaze which contains gold, a highly conductive material. The video below shows how conductive pieces can interact with a surface.

To capture touch I’ve experimented with the Adafruit CAP1188 Capacitive Touch Sensor. It has 8 channels for touch input. In this approach touching the conductive part of the porcelain is detected by the sensor. However the sensor is so sensitive that just being near the probe (black cable) is detected. This input can be used to trigger an audio event. For demonstration purposes this is just one tone.

Another way to activate the system is by using conductive objects as a switch. They close the loop and send a signal to the microcontroller. This input can be used to trigger an audio event. For demonstration purposes this is another tone.

I’ve also experimented with different clay textures and glazing. Handling the objects will create unexpected sensations and promote playful interaction.

Audio

Interacting with the objects and touching the textures will create different sounds. I’ve attached contact microphones which use a piezo element. One homemade and a Korg CM-300 BK clip-on mic. They amplify the sounds during interaction making them more intense and engaging. One of the non-functional objects has a bead inside and can act as a simple instrument. I really like the mystery of this object and want to make different variations on this interaction. The video above shows different ways of interacting with the objects to create sound.

The user will unconsciously create a soundscape while having lunch. They don’t need to achieve anything, just eat, play, relax, reflect and waste time.

Main insights

It is possible to integrate ceramics in a circuit, use different types of sensing and trigger an event.
The capacitive sensor is very sensitive. I will need expert help to fine tune this.
The contact mics attach well to the clay and amplify the sounds well. I will need to explore different ways of attaching which are more robust.

Porcelain and play

After the first explorations I decided to mostly work with the beautiful, white Jade porcelain. I was hoping for more control in creating bowls, spoons and plates. The teacher realized my fragile and soil-able work needed some extra care in handling and firing, which was very helpful.

Spoons

The spoons from the previous session all broke so I started with creating a new set. I experimented with sizes. I tried out Lotus porcelain on two pieces. It has a dark colour but is white after firing. Because the air in the studio was very hot and dry the spoons dried too fast and bend. The advice was to cover pieces with plastic to promote slow drying.

Plates

I wanted to prevent the plates from deforming and found that drying between two plasterboard was a proven way to keep the thin slabs flat.I wanted to try different shapes of plates which could provide other ways of interaction during handling. I used a wooden ring mould to press one of the slabs. I also experimented with fabrics of different weaving to create a variety of structures when using the press.

Bowls

I continued with hand building and pressing into a mould. I also used a slab with texture which I pressed into a mould. The aim was to explore different tactile experiences when handling and thus also creating different sounds. I tried to make the bowls bigger but they shrink a lot.

Playful objects

Fig. 4 Non-functional objects with different ways of glazing. Right shows the construction of the balls.

I created four non-functional objects: two discs en two balls and experimented with texture and glazing. My favourite is the ball with the ceramic bead inside. It functions as an instrument. With the help of the teacher I created a mould of half a ball. Then I put the halves together. They didn’t break. They did need a lot of sanding down. They aren’t perfectly round which serves their playful purpose very well. The discs are very abstract and can be used as one pleases. The sides have a different texture.

Technology

In designing I kept the technology in mind. Some pieces are tech free, to others I will attach a contact microphone and some are made conductive and will act as a switch or touch sensor. I also bought a silver Kintsugi set. Kintsugi is a traditional Japanese way of repairing pottery. One plate is prepared in two parts to be made conductive after firing and glazing. I will also apply this technique to one of the earlier pieces. More about this in the next blog.

Main insights

It is very challenging to glaze on both sides of the thin plates. (Fig. 2) Even though they dried flat they deform during firing because of the support props.

The shrink percentage is very high. I’ve measured values between 13 and 23 % which makes designing difficult.
Lustre runs even after taping (Fig. 2). Don’t use it in closed spaces, it gives you a headache.

Experimenting with clay

Connecting with the everyday

My design research practice focuses on making everyday activities more meaningful through interactive, embodied interaction. I explore how meaning can emerge by designing opportunities for novel connections with the objects, the activity and the body. The connections I look for are characterized by responsivity, openness, and speaking in your own voice, which enable the true transformation of both parties. This is known as resonance*.

From earlier research a set of everyday activities emerged which showed promise for resonant redesign. One of them was lunching. I didn’t find very much data on lunching but it is clear that lunch in the Netherlands is pretty bleak. It mostly consists of bread (absolute winner), soup, pasta, salad and fruit, eggs and dairy products. So the tableware should fit these products.

Fig. 1 Different plates, flattened under press with cloth. Porcelain f.l.t.r.: Limoges, Jade, Cellulain, Limoges.

I’ve started on a journey to create a set of lunch tableware which can enhance a sense of connection. They should be engaging to handle and at the same time have conductive properties to become part of an interactive system. I envision the output of the system to be sounds. The sounds can come from handling the objects and from prerecorded samples triggered by touching and moving the objects.

Working with ceramics

I have been wanting to experiment with clay and ceramics for quite some time. This and the following blog post will focus on the preliminary results of working with clay. The possibilities and limitations of working with specifically, porcelain and different conductive materials and methods for making ceramics conductive. I signed up for six classes at [Beeldhouw]AtelierBredawhere I had the freedom to work on my own project and use the knowledge and equipment available and try out different types of porcelain.

Techniques

I chose to not work with throwing. I own a set of porcelain tableware which is just perfect. I know I will not be able to surpass this and won’t add anything new by trying. Instead I wanted to explore direct manipulation and explicitly not go for the perfect cup or plate. The pieces would show the hand of the maker and traces of the techniques used. If this is done well it will add life to the objects which will make them engaging and attractive to handle. It will make them speak in their own voice. In line with the imperfect and exquisite Japanese ceramics.

Fig.2 Small bowls constructed from rolls and pressed into a shape. Partly glazed. Clay f.l.t.r.: white earthenware, Jade, Limoges, Jade.

I started out by making a plate (for sandwiches or salad), different sized bowls and spoons for stirring and eating soup. For these objects I wanted to explore the shape, size and texture. To increase engagement I wanted to add playful, unconventional elements which the users are free to use in any way they want during the lunch experience.

Fig. 3 Playful objects for displaying fruit for example. All Jade porcelain. First firing.

Conductive ceramics

I’ve made a start with making the conductive. It isn’t straightforward. I want the conductive material to be a part of the ceramic so it can be used as proper tableware. This excludes conductive tapes and paints. The first thing to try out is gold lustre. It contains gold particles and gold is a very good conductor. I’ve applied it to a plate and a bowl with varying success. However, if the lustre is opaque it works well.

Fig. 4 Two experiments with gold lustre. Left a test setup with Arduino which works. Right failed experiment.

Main insights from the first six lessons

Porcelain is very hard to work with compared to other types of clay. It is very limp and stains easily during manipulation (Fig. 2).
Thin pieces change shape during firing (Fig. 1).
Working with ceramics takes time. Drying the objects, firing and cooling down easily takes three weeks.
Things break easily. During drying, firing and handling. I can use this in my designs (Fig. 3).
The white porcelain is very prone to tainting. Nearly all the pieces showed specks or flecks (all images).
Even transparent, colourless glaze is difficult to work with. It is hard to know how much to apply and get the right fluidity (Fig.1).
Lustre is conductive. Its adhesiveness is a bit unpredictable. It’s poisonous… (Fig. 4)

*Rosa, H. (2019). Resonance (J. W. Wagner, Trans.; Kindle). Polity Press.

Performance design

The make-up performance at Damians Dressing Table aims to provide the audience with a (stylized) insight into the intimacy of the performers dressing room. For the performer the make-up performance is meant to stay connected to their own body and the (objects on) the table whilst transitioning into the drag persona Damian and preparing for the drag performance. It should also aid in establishing intimacy with the audience.

The make-up preparations consist of five distinct steps: putting on the tablecloth, combing and styling the hair, “shaving”, applying and fixing face paint (Fig. 1.) and putting on fake jewellery.

Fig. 1. Scene from the make-up performance.

The design challenge was to strengthen somaesthetic design elements in these steps to create the connections described above. How one executes the steps (e.g. pace, movements, interaction) may either support or disrupt the somaesthetic qualities embedded in the design.

Method

Fig. 2. Still from a practice round showing that is was better to insert both arm simultaneously.

The performance was designed over nine iterative rounds. Every round was recorded on video and analysed (Fig. 2.). Some observations concerned practicalities such as workflow, sound quality and preparation. Others pertained to movements of head and hands and their tempo or emphasis. Interactions with the objects evolved to make them more essential or fluent. Modifications were made in the transitions between steps. Here conscious breathing and non-functional movements provided closure.

Outcomes

The handling of the objects evolved into gentle, slow and step-wise interactions. These types of interactions create attention, care, calm, intimacy and meaning (Lenz, Diefenbach, and Hassenzahl 2013). The final performance was a natural looking routine. It was engaging and relaxing for the performer. It provided some surprising interactions for the audience to watch. This was highlighted by the set-up: a camera was placed above the table and provided a live stream of the interactions from the perspective of the performer (Fig. 3.).

Fig. 3. Performance with live camera. Image Sas Schilten.

Looking back the performance would have gained naturalness and ease if it was practised more often. This would have made it more embodied for the performer thus contributing to its expressiveness.

References

Lenz, Eva, Sarah Diefenbach, and Marc Hassenzahl. 2013. “Exploring Relationships between Interaction Attributes and Experience.” Proceedings of the 6th International Conference on Designing Pleasurable Products and Interfaces, DPPI 2013, 126–35. https://doi.org/10.1145/2513506.2513520.

Touch to Sound

As explained in a previous post, touch triggers sound. The aim was to evoke a sense of surprise, presence, and alertness in the performer and the audience.

The first exploration was on the types of sound to use. Although surprise was one of the aims there should be logic to the sounds. In the early stages of the design, I experimented with sounds generated by objects on the table such as bowls.

The sounds were interesting but they required interaction that wasn’t in line with the make-up performance. There weren’t many bowls on the table and attaching these sounds to other objects didn’t make sense.

The most logical was to let the three beauty products (hair cream, setting spray and tissue) respond to touch. Touching them was a natural part of the performance. I decided to use the four starting notes of the song Damian would perform after the make-up performance: I put a spell on you, performed by Screaming Jay Hawkins. The notes are very characteristic and fit well with the purpose of the sounds.

With musician Hans de Zwart I explored different atmospheres using a Roland JV-1080 synthesizer with the “World” expansion card. The sounds varied between interesting, bland and bizarre.

We settled for a subtle but fascinating sound. It affirms the action and gives a sense of presence to the interaction.

Mirror and Belt

This post describes the experiments for creating a mirror that aims to assist in regulating breathing by both visual and auditory cues.

The mirror acts as a device for entrainment, setting the pace for a slow breathing rhythm. Five bright LED lights (Neopixels) placed around the mirror fade in and out for five seconds respectively (Fig. 1.). Research has shown that slow breathing aids relaxation [1]. However, the mirror with the lights also had the practical function of illuminating the face during the make-up process. So the dimming of the lights couldn’t be too low but still bright enough to see the difference. The lights are dimmed to around 30% and go up to 100% brightness. This works well with moderate daylight and average-strength artificial lights. The bright lights in the exhibition space made the changes hard to see.

To aid the entrainment process audio feedback was used as a subtle noise stream generated by a piezo element placed on a belt. This was worn around the waist to capture the belly movements. The belt contained a piece of fabric which rubbed against the piezo element and created soft noises which made audible the different phases of the breathing process.

Fig. 2. Test setup with belt, recorder and fabric samples.

The sounds should be relaxing and informative. 13 different fabrics were tested. They ranged from very soft fleece to stiff felt. They were rated on a one to three scale on how well they represented the breath flow and their pleasantness.

The top six noises were recorded and assessed using a surround speaker set (Fig.2.). The final choice fell on a stretch fabric folded in four layers (Fig. 2. white fabric on the right). It made all the phases of the breathing process, from the very beginning to the end, audible in a pleasant way.

For the performance, an expansion filter from the mixer panel was applied to enhance the sound.

[1] Zaccaro, Andrea, Andrea Piarulli, Marco Laurino, Erika Garbella, Danilo Menicucci, Bruno Neri, and Angelo Gemignani. 2018. ‘How Breath-Control Can Change Your Life: A Systematic Review on Psycho-Physiological Correlates of Slow Breathing’. Frontiers in Human Neuroscience 12 (September): 1–16. https://doi.org/10.3389/fnhum.2018.00353

Table interactions

The design question for this experiment was how interaction with the table and the objects enables the make-up practice and also enhances the connection with the environment (table and objects) and decreases the sense of self by inducing wonder or awe. The latter are manifestations of self-transcendent experiences [1].

Fig. 1. Different prototypes for interacting with the objects.

The layout of the table was explored earlier. This post describes the actual interaction with the objects and the exploration of estranging elements [2] which may enhance surprising engagement and intimacy with the objects. The position of the cables towards the person provides a sense of intimacy and logic [a]. An outline around the objects has a calming effect because everything seems in its proper place [b].

Touching the objects should spark some wonder. The first attempt was to use a “magic wand” [c]. Touching connected objects with the wand would start a sound sample and cause surprise. The Bare Conductive Touch Board [3] was used to play pre-recorded samples. Objects are connected with conductive yarn to the board when the wire is touched by a hand or a conductive object the appropriate file plays.

But there was no logic in introducing an extra object like a wand. By making the beauty products conductive the actor can simply use their hands. Conductive tape was applied to the products in a closed circuit [d]. This way the tape can transport the touch via conductive yarn to the Touch Board. Fleece fabric isolates the tape from the yarn. The connection is made when some downward pressure is applied and the space between the tape and yarn is reduced.

When applying face paint water is needed and the paint stains the hands which is very uncomfortable. So a bowl of water and a napkin were added to the prototype [e]. Transparent plastic was placed under the wet parts to prevent staining.

The final prototype supports the practical use of all the objects and adds an element of wonder and engagement when the beauty products are touched.

References

[1] Yaden, D. B., Haidt, J., Hood, R. W., Jr., Vago, D. R., & Newberg, A. B. 2017. The Varieties of Self-Transcendent Experience. Review of General Psychology. Advance online publication. http://dx.doi.org/10.1037/gpr0000102

[2] Danielle Wilde, Anna Vallgårda, and Oscar Tomico. 2017. Embodied Design Ideation Methods: Analysing the Power of Estrangement. In Proceedings of the 2017 CHI Conference on Human Factors in Computing Systems (CHI ’17). Association for Computing Machinery, New York, NY, USA, 5158–5170. https://doi.org/10.1145/3025453.3025873

[3] Bare Conductive. 2025. Touch Board. https://www.bareconductive.com/collections/touch-board Retrieved: 12-2-2025