Long before anyone considered using a computer, a parlor game has been playing at Cantonese brunch tables for years. Someone at the table, usually a relative or an overconfident college buddy, comments that your preference for siu mai over har gow reveals something profound about your character as soon as you settle in and the carts begin to roll.
They’ll call you basic, or secretly an introvert. It’s absurd, of course. However, it’s the kind of drivel that sticks because people want their food to have a purpose. Strangely, artificial intelligence is now picking up the same thread, and it may not be completely incorrect.
| Topic | AI Taste Mapping & Dim Sum Personality Profiling |
|---|---|
| Field | Artificial Intelligence / Food Science / Consumer Psychology |
| Key Research Area | Cross-Modal Correspondences & Food Preference Prediction |
| Notable Researcher | Carlos Velasco, BI Norwegian Business School, Oslo |
| Supporting Research | Charles Spence, Cross-Modal Research Laboratory, University of Oxford |
| AI Models Studied | ChatGPT-4o, ChatGPT-3.5, Google Gemini |
| Culinary Reference | Sandy Shi, Executive Dim Sum Chef, Wynn Las Vegas (Michelin-starred Wing Lei) |
| Core Technology | Machine Learning, Collaborative Filtering, Behavioral Pattern Analysis |
| Applications | Food delivery apps, restaurant personalization, consumer profiling |
| Reference | BBC – AI Started ‘Tasting’ Colours and Shapes |
What some foodtech experts refer to as AI taste mapping is a straightforward but somewhat frightening concept. After being trained on millions of user interactions across meal delivery platforms, machine learning systems are becoming impressively adept at not just forecasting your next order but also deriving behavioral and personalityrelated patterns from your past selections.
When you often choose taro puffs when dining in groups and soup dumplings on rainy evenings, an algorithm not only recognizes the pattern but also begins to create a profile. Carlos Velasco of the BI Norwegian Business School in Oslo has described it as a sort of digital taste fingerprint, but he was referring to flavor profiles more generally.
It would be easy to write this off as marketing gibberish masquerading as data science. And perhaps a portion of it is. Beneath the hype, however, is a genuine body of study that extends farther than most people are aware. Research conducted in the 1970s demonstrated that people identify particular colors with particular flavors, such as brown with bitterness and pink with sweetness.
These correlations have been repeated across cultural boundaries. More recently, Velasco and Charles Spence at Oxford discovered that similar crossmodal correspondences are surprisingly reliably reproduced by generative AI models when prompted. It turns out that ChatGPT4o tastes color quite similarly to humans. Although the consistency is difficult to completely deny, it’s possible that the machine is simply repeating known findings.
What happens, therefore, when you apply this reasoning to dim sum, a dish that is essentially designed to reflect personality? Think about the spread: you have the crowdpleasing, safe siu mai, the slightly daring har gow, the textural surprise of a taro puff, and then, like a dare, chicken feet sitting on the table.
For more than 25 years, Sandy Shi, the Michelinstarred dim sum chef at Wynn Las Vegas, has observed how patrons place orders. She has observed that American diners take a different approach to the food than Hong Kong locals, concentrating more on the food itself and less on the social custom. There’s a feeling that the first thing a person goes for when the cart passes by says something about how comfortable they are taking risks, how receptive they are to new textures, and perhaps even how they respond to peer pressure.
These kinds of signals are already being tracked by AI meal recommendation systems. Not only do websites like Uber Eats and DoorDash record your orders, but they also record what you hover over and skip, when you explore, and what you leave in your cart. You are grouped with users who have similar preferences through collaborative filtering.
Deep learning algorithms identify patterns that you might not be aware of, such as your consistent avoidance of fried foods or your preference for foods with similar textures across different cuisines. The algorithm is creating a mosaic of little choices that, when combined, begin to resemble a portrait rather than reading your thoughts.
It remains to be seen if that picture truly conveys anything significant about your character. The you are what you eat school of pop psychology has a long and rather embarrassing history of attempting to infer character from dietary habits. The majority of it fails to withstand close examination. However, the AI version differs in at least one way: it doesn’t make lofty assertions about your soul.
It is estimating your next order using probability, and those estimates are becoming painfully accurate. The system does not assume you are an introvert if you consistently choose the char siu bao (steamed, not baked). It’s important to note that you have similar behavioral tendencies to a group of users who also frequently place latenight, solitary comfort food orders. Humans add layers of interpretation to their personalities.
Even so, it’s difficult to ignore how well the two work together. Machine learning finds behavioral clusters that closely correspond to the folklore at the dim sum table, such as the tofu skin roll person marching to their own beat and the xiaolongbao lover craving surprise. Not entirely, and most definitely not in a scientific manner. However, it has a rhyme. According to Spence, AI may someday produce untested theories regarding crossmodal associations that may be tested on actual individuals. It turns out that food is an unexpectedly rich signal.
Whether AI can match your personality to your taste is not the true question. In a strict sense, it most likely cannot. The question is whether the patterns it discovers the subtle regularities in your eating habits, when you eat, and what you grab for when no one is around tell a compelling tale. Perhaps be mindful of what you grab first the next time the cart comes by.
i) https://www.carlykan.com/blog/what-your-favorite-dim-sum-says-about-you
ii) https://www.ai4lifecoach.com/ai-food-preference-prediction-recommend-foods/
iii) https://www.buzzfeed.com/michellerennex/yum-cha-dim-sum-quiz-qualities
iv) https://www.bbc.com/future/article/20241220-an-ai-started-tasting-colours-and-shapes-that-is-more-human-than-you-might-think