TastePrint：味の分布を設計する3Dフードプリンタ

Welcome to Data Skeptic Human-Computer Interaction. Today on the show I'm speaking with Yamato Mayatik, a PhD student working at the intersection of HCI and digital food fabrication. What is digital food fabrication? Well, it somewhat speaks for itself and we'll get into the details, but basically it's printing edible artifacts. It's not quite the replicator from Star Trek The Next Generation, at least not yet. But Yamato's project TastePrint is quite an advancement. It's a 3D food printing system that controls where taste sits inside the food. And not just its shape. Food science knows materials, and HCI knows how to design interactions to stimulate creativity. I think TastePrint can be a bridge between these two. Let's get right into it. I'm Yamato Miyatake, a PhD student at Saitama University,

working at the intersection of human-computer interaction and digital food fabrication. Could you share a few details on what digital food fabrication involves? So digital food fabrication involves creating edible materials using digital fabrication tools, such as laser cutters, 3D printers, or some computationally designed cycle molds. And in general, is there a goal to, like, I don't know, make a new consumer product? What does the field kind of have on the horizon? What are some of the long-term research goals? The research goal will be making the cooking process computational. So if we can design some cooking recipes or cooking process inside computer or computationally, we can employ some optimization techniques and also we can create new dining experiences through the power of computational technologies.

Are you familiar with the term gastroentomology? Sorry, gastroentomology. So I may have actually said it wrong, but there's a type of restaurant, and I've been to one, where they try and involve science and food. So like there was some appetizer that they brought with dry ice, and, you know, it had a foam and this kind of thing. And it was sort of the trappings of science in a dinner. Is this a related idea? Yeah, they are kind of related. So those fine dining ideas are also related to our domains. But our main goal is not only making some appealing food itself, but the more new dining experience that can be deployed in the normal daily dining environment. So I am a bit of a home cook. I work from home and I take some time every day to make a lunch. And that time is valuable.

I could have deployed it professionally, you know, and charged some clients the money instead, but I chose to, I guess, prepare the meal. If I had a robot in my kitchen, I could have just pushed the button and it could have made the food and I could have done my role. Is that sort of within the domain of your goals? Well, that is one of the directions. But I'm more interested in how people collaborate to work with those robots to stimulate their cooking creativity. I think those robots or those systems will not take the people who are cooking creativity. People can create more creative dishes using those robots instead. Could you define the taste print system? I know we haven't really talked about it exactly yet, but that's, I guess, central to the work that I've learned about from reading some of your papers. What is taste print? So TastePrint is a new 3D food printing system that can control where the taste is printed, placed inside a 3D printed food.

So a conventional food printer can shape food layer by layer, but the taste is often uniform. Our system exudes one base food material and then uses programmable sprays to spray liquid seasoning at selected positions after each layer. So as the consumer, if I was eating one of the products that TastePrint produced, what is my experience and how is it shaped by this process? So TastePrint can create distribution for some taste button inside food. So we are imagining that people can experience some temporal difference of taste. For example, the people will feel some stronger salty taste first and then feel more sweeter taste later. Those kind of taste distribution, this pattern can be created. Is this in line with what traditional chefs do? I've never made my own recipe. I'm curious if chefs think in this way that there's an outer taste and an inner taste

or if this is something new that work like yours is unlocking. The taste distribution, a special taste composition itself, is not a new biology. The professional chefs are already employing those methods already. But what is different from those traditional methods is the reproducibility. We are using 3D food printer that enables us to design the 3D model, the shape of the food, and also the taste print enables to design the taste distribution inside the computer and it can be reproducible. A little difference. So let's say I was someone who was going to become a user of TastePrint, and I had some idea for, I don't know, a new food product or something like that. How do I work with the system? How do I configure it? So users can just specify the taste distribution and present a graphical user interface, which is quite similar to the conventional slicer for a 3D printer. So people start loading the 3D model into the system,

then the system slices the model to each layer based on the printer's configuration. And then in addition to the shape customization, users can specify where the taste to be placed and the amount, intensity, so on. And the system automatically calculates those taste configurations into the machine commands such as G So I guess my interpretation is that TastePrint is a robust system for specifying how to create a thing Is that correct? Yeah, that's correct. It can robustly create a taste distribution, yeah. So then I think of it more as a tool that I can configure. If I want to do the optimal configuration, or I don't even know what I'm optimizing for, I guess, what are some of the interesting scientific questions we can ask with this tool? Yeah, that's correct. The taste brain itself is a tool to create a robust taste distribution. By using the tool, we can create complex taste distribution

and we can investigate how people will perceive against the complex or different taste distribution. That is, I think, yet unknown in the perception domain. And some food research have shown that non-uniform taste would create different taste perception against uniform taste distribution. But we don't know which taste distribution would make what kind of taste perception. We can do new experiments for those new taste patterns. And also we can think about how to optimize the taste pattern to create ideal taste perception. And I'm curious if you have thoughts on how research like yours can touch the everyday person's life. I'll take, for example, sweeteners. There are low-calorie or zero-calorie sweeteners. I was once someone who drank a lot of regular soda, and I switched to diet, and that was a good thing for me, actually. I was able to lose weight through that process. So thanks to chemistry, this happened for me.

Does your work kind of intersect with these dietary optimization kind of goals along those lines? That is actually out of the scope of the current paper, but that is the long-term goal of our study. The taste print, of course, combined with some aroma or more different ingredients instead of just sweetness taste. Yeah, this is different to be our long-term goal. So with a combination of aroma or sweetness aroma and low sweetness ingredients, we can definitely optimize the taste distribution to reduce the total amount of sweetness component while maintaining the intensity of sweetness perception, which definitely will be helpful for healthcare. Well, when I think of my eyes, they see light, obviously, and I think of it in terms of red, green, and blue as the two dimensions, all the colors I see. I don't necessarily know the dimensions for my taste buds.

And as you point out, aroma is this sort of almost related field coming in that I didn't expect, that my sense of smell is impacting my sense of taste. Is there a good science around how we represent mathematically the notion of taste? Actually, I think there is no specific mathematical model for representing taste, because taste perception is quite complex perception and also known as multi-sensory perception, combining eye information, vision information, smell information, and also taste, tastebot information. We are perceiving the taste or flavor through the combination of those information, and that is not defined by only through TasteBot, actually. So a complicated and interesting thing to study for sure. Yeah. Well, let's get back to TastePrint. I know you did some deployments where there were people using it, I think, in their homes or at least in a research environment. Can you talk a little bit about the experiment you conducted?

We conducted two types of experiments. One is for hardware and the other is more user experience and user study. For hardware, we conducted, we experiment, we experiment the accuracy of the spraining pattern. And for user study, we conducted two user studies. One is the actual tasting experiment. The other is formative design experiment. And through user study, through the tasting experiment, through the user study, we found that, so we conducted initial user study and gained some feedback from users. But the result is still exploratory, and we couldn't say the test distribution or usability of the system is perceivable or usable at this pace. So more further logical user study would be needed to show the user experience. Yeah, I guess my interpretation of that is that there's an axis in all of science, and this is something early stages that humankind is getting to.

And are you familiar with the TV show Star Trek The Next Generation? Actually, no. Okay, well, you might want to look it up on YouTube. It's a science fiction program, and it takes place on a spaceship, and they have a technology called the replicator. And you can go to the replicator, and you'd say, I'd like two scrambled eggs and hash browns, and it sort of magically appears, or scientifically appears. So it's a machine that effectively 3D prints food, but it does it very quickly in terms of the show, and it can produce whatever you would like, which I guess is sort of maybe the ultimate endpoint of a journey like the research path you're on. If you could give a large language model a prompt of what food you want, and it could manifest the food. This is sort of like, I guess, the perfect outcome, which we're somewhat distant from. So on a journey from maybe the current state of technology to the realization of that science fiction idea, Where do you think the current state of science is?

We are currently thinking to personalize the taste of our food using those tools. A taste print can create robust and reproducible taste with different taste ratio or components. If we could get some feedback from users about their preference, perception, or feeling, or binary feedback, or language-based feedback, we can train AI models to learn one's perception or one's preference and then create ideal fruit that is perfect for one's preference or health condition. You mentioned that taste print is robust and reproducible Could you talk a little bit about how other labs might adopt the process to you know do science following your procedure and your techniques Yeah, I think taste print is an initial approach to separate the formulation of shape of the food

and seasoning during printing. So a conventional way to create multiple or varying taste during printing is to employ multiple food inks loaded into the 3D food printer. So I believe other scientists can follow this direction. I think it shouldn't be limited to spray-based seasoning method. They can also combine other powder-based mixing or other methods to create valuable taste. But I definitely think separating the form printing and seasoning should be one promising direction for future 3D food printing and food researches. And could you share some details on the feedback you got from the people that participated in your research study? For taste evaluation, from the taste evaluation study, the feedback was actually varied across participants. We conducted a testing study with uniform taste distribution and centralized taste distribution,

and some realize that the taste distribution is different across the two food materials, but others distinguish it all in then. So we have to investigate it further. If it is caused by the intensity of the taste itself because of the participant group, because the conducted user study was still small and limited, we need further investigation for that. I guess the human taste bud system, I don't know a lot about it. I know my eyes are sort of limited that if I go out and buy the best TV I can purchase, it probably has a finer grain resolution than my eyes can even appreciate biologically. So I think I have a sense that technology in terms of view screens has gotten past where we are as human beings biologically. Taste-wise, do we have a similar understanding or a good way of expressing how much flavor we put into something to generate the experience we're looking for?

There are a bunch of researchers investigating the humans, for example, taste resolutions or taste discrimination towards some basic tastes. But for those evaluations, we need professional evaluators, actually. And I think it is quite difficult to evaluate the taste of casual users, actually. That is one of the difficult points to conduct the user study. It is difficult to collect the professional taste evaluators. But even if we collect those professional evaluators, our main focus is the casual users. So there is a distinction between the evaluators and actual users. So we have to create some effective or better user study systems to get more informative information from users regarding the taste and variations. Naturally, having more participants would be better. It's a bigger sample. Bigger sample is always better.

But are there other dimensions you would want to improve upon? For example, like maybe your equipment had only a certain limited resolution and you'd like to spray with a finer grain resolution. Is there any technology limits you'd like to see surpassed to get greater insight? Yeah, the resolution is one of the important limitations in our setup. We are currently using a simple spray and the resolution is not quite high. But there is also a threat of increasing the resolution would increase the time to print, and we are still not sure how much resolution is needed for effective user perception, taste perception. Another limitation is the material cross effect. We so far use mashed potato for experimental material. The mashed potato is very useful for laboratory-based experiments, but we also need to investigate the effect of absorption or dissolution.

When we employ other materials, additional calibration would be differently needed for other materials. Are there any other materials you in particular would be interested to explore? For example, vegetable puree or cookie dough would be an interesting material to try. But there's one limitation on 3D food printer. the 3D food printer can only employ the slightly pasted material so far. Well, in measuring taste perception, you can present food to a human, and they can consume it, but they have to give you something back. How do you solicit the right feedback from them? I think there's multiple ways to get feedbacks from users. But I think for casual users, it is difficult to specify detailed test feedback because understanding the taste perception is a difficult domain to describe. So we are currently working on preference learning based on the user's taste feedback.

In the study, users just give the binary feedback. For example, if they like taste A or taste B, that is a more simpler feedback, and they can handle. That is one of the possible directions for gaining the user's feedback. And what are some of the open research questions you'd like to explore? Open research question would be creating a closed loop using the test print system. So closed loop means learning users' perception or preference based on their feedback. So test print can create variable taste, but we still don't know what pattern would be effective for users. And the reverse design using the perception model to create the taste pattern is also still difficult because, as we said, the taste perception cannot be solved analytically. We need learning model and closed loop to learn research perception and create a real taste pattern.

And yeah creating those model and closed loop would be different to future research goals The other project I wanted to ask you about is ChewTect Could you talk a little bit about what the project all involving Yeah ChewTect is a project to create temporal texture perception. So in food research domain, it is known that the texture perception in the first bite and progressive subsequent chewing phase is perceived differently. But the recent 3D food printing research has been investigated one uniform texture perception. So, Tutect explored the two texture perceptions differently and found that by tuning the internal structure of the food, we can slightly independently tune the texture perception for the first bite and subsequent chewing texture. Very interesting. Could you expand upon that? I understand, I think,

on some level of what you mean by like the initial and then the chewing, I'm thinking of the snap of a hot dog or a sausage that there's a bite and then it's chewy or like a creme brulee where there's a sugary top I have to crack into a soft thing. Is this the two stages you're talking about? Yeah, that's the one. Yeah. So we feel differently for the first bite, like crunchy texture, and also the subsequent chewing food like biting and crushing the food. Is that something you can express in a 3D printed food? In the study, we created food using the 3D printed silicone molds. So it might be not good to say the food is 3D printed food because it is created by 3D printed molds. But we can also think create the food using 3D food printers and combining those technologies with taste print would be different, really interesting direction, I think.

I guess a lot of this, I would assume, is about scaling it at an industrial scale, that when you find a really optimal design or a delivery system for just the right crispness or first and second bite thing, that you then perfect that into a machine that produces an infinite number of the perfected unit. Do you see it going in that direction or do you have a different vision? for it? Industrial level or mass production in those, I think they can employ more precise or expensive machines in those domains. And I think the 3D food printers or taste print can be used in more end users for the concept stage to create a different new taste experience or new food. I think it's a different direction to go. So do you expect that within our lifetimes, we will probably have a 3D printer in our kitchens in the future? Ultimately, but I think one first step is to deploy the 3D food printer inside some professional kitchen environment.

That makes sense. Is that being done or is that still arm's length out? For proof of concept, we are currently collaborating with a kitchen and talking how to deploy those printers in an actual restaurant. And I think in the world, there are some restaurants that are employing 3D food printers already. Very cool. Exciting times. The topic for our season on DataSkeptic is HCI, human-computer interaction. Could you contrast or relate how that works or how that aligns with your own research? So Taste Print is a new 3D food printer And in food science there are many researches 3D food printers But I think how people will use the printer And how we can change the people's dining experience Is not yet fully explored So I think the Taste Print can be a bridge To connect the food science and human-computer interaction There's many scientific knowledge

on how to design something or how stimulate people's creativity. And we have a bunch of knowledge on that. And we can bring those knowledge to the food science. And we can also get some feedback from food science. And then we can go for more. Those collaborations effectively create future better research direction and in-the-world environment deployment. Well, let's imagine there is someone listening right now who is a chef. They're a little avant-garde. They're interested in topics like this. And they also have a friend that's technical who can read a paper like yours and look at the designs and roll it out. What sort of opportunity is there? And what kind of startup cost do you think they need to maybe launch a little boutique restaurant that has some 3D printed food for people to come in and check out? We definitely need collaboration between technologists or scientists and the chefs or professional practitioners.

And in this stage, it is not yet off-the-shelf technology. We need collaboration and we need to think how to adapt to the materials they want to use or how to use the technology in the world environment. Those collaboration and cooperation from both directions will be needed. and it will create more better collaboration and better dining environment experience. I'm interested in seeing those kind of new restaurants, and I hope my technology or future research will be employed in those kitchens to create new food experience and dining experience. Is there anywhere listeners can follow you online? The listeners can find me on Twitter, on X, And if you read the paper, please contact me through the email. Very cool. We'll have some links in the show notes for listeners to follow up. And then I always like to ask, what's next for you? My next research direction is, as I said, the human in the loop,

a closer loop, preference learning. Yeah, we're actually already working on that, and I hope to make it public and all that. Well, sounds exciting. Definitely keep us in the loop when the publication comes out. Ivano, it's been great talking to you thank you so much for taking the time to come on and share your work thank you for inviting me