→ ① Preface(EN)
→ * 序言(CN)
→ ② Translations 译文
→ ③ View 观点
→ ④ Project 项目
→ ⑤ INUN Tabloid 氤氲小报
→ ⑥ Miscellany 杂记
Editor's Note:
This is the first installment of the "View" series presented in interview format by the creatively limited INUN. The series is divided into three parts, and from the perspective of design researchers, we engage in discussions with cognitive neuroscientists on topics related to the application of neuroscience research in design practice.
The design process is filled with intentionality and a sense of cybernetics. Simply put, it often carries an optimism of controlling and transforming the world, whether blindly or rationally. Faced with the unknown world, the inexhaustible curiosity arising from the uncontrollable sense of insecurity further compels designers to eagerly await at the doorstep of other disciplines or impatiently push open the doors, shouting—what are you doing? This allows designers to have an additional set of tools to operate, rather than simply appropriating dusty classics. Recently, we've observed that the gateway to the field of cognitive neuroscience is crowded with designers and their research teams. Some have begun to use these tools, attempting to refine a set of operational research methods within the field of design.
Intuitively, such anticipation seems worthwhile. We have been exploring the path of "cognition" for too long, and a considerable amount of discussions about "how design can make the world better" all point to a core question: How do we perceive the world? With technological support, can we design a comprehensive experimental mechanism to transform so-called sensory experiences, evaluations, and designs into data, charts, and discourse?
However, science tells us that intuition cannot serve as the sole argument. After seeing numerous experiments in the design field and standing once again at the gateway of cognitive neuroscience, we want to clarify a simple viewpoint: how do these experiments in design look from the perspective of real cognitive science? We interviewed HDC, a scholar researching visual science and experimental psychology at the University of California, Berkeley. In the interview, we discussed the feasibility of experimental design in design studies, the basic experimental design process in the field of neuroscience, and a scientist's contemplation on "beauty."
Part 1: The Feasibility of Design Experiments
INUN: We've noticed that many designers are now using brainwaves to evaluate the quality of existing designs. Some are even using these experiments and devices for design predictions. So, from your professional perspective, is it feasible to put a device on someone, generate a set of data, and use it to evaluate the design of a building space, a street, and similar experiments?
HDC: In reality, experimental data is quite complex. Once you have the data, how you process it depends on the questions you have (what kind of results you want from these brainwave signals). The way you ask the question determines how you handle the data. So, it's not possible to use a very rough question and expect the experiment to give you precise results. There's also an issue with the analysis methods in these design experiments. The analysis tools they use are quite limited. I looked at the software used in the experiments you mentioned, and the results are basically a spectrum that corresponds to the active regions of the human brain. But analyzing it solely based on the spectrum is not sufficient.
INUN: Do you think the research ideas behind these experiments are feasible?
HDC: I think the idea is good because design itself is something centered around people. I don't know much about design theory specifically, but I believe that if you design something, it stems from the creator's creativity or ideas. For example, a public sculpture or a building is meant to be there for everyone to see and perceive, or if you design a landscape that thousands of people experience every day. If you, as a designer, didn't anticipate that this thing might evoke some emotional changes in people, for instance, if the landscape feels oppressive, it may not be good. Or, if you intentionally designed something oppressive to convey a certain message, your audience might perceive it differently, and they might not necessarily grasp your intended message because emotions are inherently subjective.
From the perspective of cognitive neuroscience, changes in emotions are largely unconscious. You don't actively generate your emotions; they arise passively when you encounter things. Everyone can feel this; the changes in emotions are related to your memories and experiences. People with different experiences, when in the same situation and exposed to the same stimulus, will have different emotional responses.
INUN: So, can we understand it like this: for example, if someone conducts experiments using changes in brainwaves to evaluate the quality of a design, whether it's landscape design or urban design, is this akin to using a rational tool to evaluate something emotional? But, as you mentioned earlier, emotions are random and not under your control, so how do you evaluate and quantify them? Does brainwave analysis provide us with such a tool?
HDC: It might not be that simple and direct. First of all, in conducting EEG experiments, there are several crucial things. The first point is that we need to establish the correlation between events and signals. EEG signals are very noisy because we are dynamic beings. Even in a state of sleep, our brains remain highly active, including both conscious and unconscious activity. All brain regions are constantly active and interacting, and signals collected at any given moment are particularly noisy. So, if you want to obtain a signal and claim that this signal is caused by a specific event, you need to provide sufficient evidence to prove that the signal is indeed caused by the desired event; otherwise, there's no way to demonstrate any correlation. EEG signals also have a delay. In an experiment, from the moment I present something, the information enters your eyes, and eventually, your visual system starts processing this information at the back of your head. There's a delay of roughly 50 to several hundred milliseconds in between. Therefore, you need to observe the signal fluctuations in a specific brain region at some time point (zero time) after the event occurs, which could be one to several hundred milliseconds. Regardless of the analysis method used, you need to associate the signal with the event.
To achieve this association, you must repeat the same event many times because, as we mentioned earlier, the noise is very high. For instance, if I want to show you an object to test the brain's response, I may need to repeat this action 100 times. After that, I lay out the signals obtained over time, using the time when the event occurred as the zero point, and then average them. Averaging helps remove noise, allowing you to obtain a relatively clean signal. Based on what I've seen in the experiments you showed me, the degree of noise reduction in the signal seems insufficient.
The second point is that a large amount of data from different participants is needed—a relatively large sample size for the experiment. As we just discussed, everyone's experience is quite different. If you want to understand a very general neural phenomenon, you need to test more people to link your experimental results with correlations.
Because if three people do the same experiment and get three different results, it's obviously a weak correlation. Let's assume that conducting an experiment is a simple process from A to B to C: A is the event, B is the signal, and C is your interpretation of the signal.
The experiment you showed me currently has a question mark in the A-to-B process. How do you prove the correlation of the experimental data? Even if we assume that the A-to-B process is flawless, and the signal is stable and universal, how do you arrive at C, which is the experimental conclusion, through what kind of analysis?
Even if you have mapped A to C, the relationship between B and C may not be accurate. The interpretation of experimental results is built on numerous previous experiments, and innovative experiments require sufficient data to support the analysis process. In many cutting-edge fields, debates are common, and conflicting phenomena are widespread. This is currently a practical problem. In the current development stage of neuroscience, our understanding of the brain is still relatively superficial. Different experimental results point to different explanations for the same phenomenon. In these examples, after processing the data from A to B, if there are issues with the design of your experiment, the signal you obtain may not be related to your experiment at all.
Exactly, this creates a situation where the foundation might be flawed from the beginning. You assume the experiment is effective, and you derive a conclusion. Everything in between may be sound, but if the design of your experiment is not aligned with your research question, the entire endeavor may be invalidated. Of course, this depends on how specific your research question is.
INUN: So, we understand that experiments like the ones mentioned earlier involve many flaws in the A-B-C process, and there might even be considerations of D: the practical application level, making it even more challenging to explain the authenticity of the experiments.
HDC: But you can't say these things are bad. Because if everyone follows the rules, it might be challenging to come up with innovative ideas. There's always someone who, when trying to create something entirely novel, is initially constrained by familiar and habitual experimental conditions. At this point, the experiment is an open-ended attempt. I think if there is value in it, then it is authentic. But I don't know if the conclusions drawn from the equipment are as good as they are presented. Generally, we need manual data processing, but in the examples I've seen, the analysis of experimental data seems too dependent on software.
Regarding applications, the use of these technologies itself poses problems. The authenticity and repeatability in the experimental stage are challenging to verify. For example, in current neuroscience research, the most popular approach does not necessarily mean it's the best. However, currently, the experiments most likely to be accepted by high-impact journals are those conducted using functional magnetic resonance imaging (fMRI).
The brain is highly structured, with different regions responsible for different functions. Functional magnetic resonance imaging (fMRI) operates on the principle that when a region of the brain is active, it consumes more energy, leading to changes in blood oxygen levels and metabolism. This process involves more iron ions from hemoglobin binding with oxygen, forming more magnetically active oxygenated hemoglobin. The magnetic properties of this region differ from those of other regions. Therefore, functional MRI generates a strong magnetic field to test how different brain regions respond to this field and to identify which area is more active. When the magnetic field is strong enough, its spatial accuracy can reach a few millimeters, capturing the response of neurons in a specific brain region.
However, it introduces challenges. During experiments, even a slight movement by a participant lying in the MRI machine can have a significant impact on the signal. If the MRI detection region is not precisely aligned, minor movements might not matter. However, due to the high precision of fMRI, even a slight movement can introduce errors into the results. These challenges add complexity to research work, and many experimental errors are uncontrollable. This is a technical issue, and because of it, a series of data processing steps are necessary to correct for the signal's errors. Yet, in this series of processing steps, it's highly possible that the true signal has already been lost.
INUN: Is it because the human brain is complex that it's challenging to obtain consistent inputs? As you mentioned earlier, even a minor movement can significantly alter the data, making it difficult for these experiments to establish a strong correspondence between the data and the final conclusion.
HDC: It's not necessarily that it's very difficult, but when conducting neuroimaging experiments, your experimental design needs to be very meticulous. Every aspect of your design process needs to be well thought out. First, the experimental environment should be carefully considered because the human brain is always active. Even someone yelling nearby during an experiment can affect brain signals. Therefore, when conducting experiments, you need to create a closed, quiet, and stable environment. Sometimes, when a person raises their arm during an experiment, a signal is immediately generated. So, you should minimize unnecessary body movements. When there are movements, you need to know how to isolate them from the data.
In reality, many experiments involve numerous confounding variables. For example, in an experiment where participants are required to click a mouse every time they see an image, you need to determine whether the obtained signal is related to the person seeing the image or to the person clicking the mouse.
INUN: After data processing, how do you arrive at C: the interpretation of the signal?
HDC: The most straightforward conclusion is that the activity in a specific brain region represents a phenomenon. As I mentioned earlier, if the occipital region becomes active, it indicates activity in the primary visual cortex, and this is undeniable. Once experiments have been repeated across various individuals, you can confidently say that this is a confirmed phenomenon.
The interpretation of EEG signals generally involves two dimensions. As an electric signal that changes over time, it essentially has two dimensions: one from the frequency domain and one from the time domain. Typically, a description of an EEG signal involves stating, after an event occurs, the signal in a specific brain region, within which frequency spectrum, begins to rise or fall, and at what point in time it reaches its peak.
So, I'm curious about your field of design. When you conduct these experiments with EEG data, what specific question are you trying to answer?
INUN: Actually, we want to use a method to quantify the perception of space. From an emotional perspective, we're exploring whether EEG can serve as a quantifiable way to assess space and evaluate it. After quantification, architects and urban planners can make improvements based on these quantified indicators. Then, we can measure these quantified indicators to see if there are any changes. We believe that what they want to do the most is, if they can quantify abstract or subjective elements and see changes, they want to confirm these changes in a rational way.
HDC: The simplest way to judge whether I feel comfortable in a space is to give me a form with a scale from 0 (uncomfortable) to 5 (very comfortable). Then, I mark my choice on the 0 to 5 scale, and you have a result that's quite sufficient. The rational confirmation you mentioned is based on an assumption that the brain signals won't deceive, right?
When people fill out a form, it involves multi-layered processing. For example, my brain might think it's a 3, but I might mark 5. There's a black box here, but in reality, neuroimaging is another black box within that black box. The problem you're trying to solve is, you think the human brain is a black box, and you want to open it: What is its perception of this thing? But when you measure my EEG, you're measuring a small black box inside my big black box. Because you still need to explain what my EEG signal represents. It's just that you've turned my subjective interpretation into your subjective interpretation, or perhaps the interpretation of the academic community.
So, if you want to conduct this experiment and your goal is to measure a direct brain signal that reflects a neural phenomenon, I think you can certainly do that. But if your goal is simply to evaluate the goodness or badness of architectural space itself, it might not be necessary.
Thanks to the interviewee HDC for providing the interview information and verification support!
Interview: Wenzhuo Cai
Editors: Lydia Li, Wenzhuo Cai, Yuanlong Zhu