每學期開學，我班上一定有小朋友問我這個問題。有時候是他們自己問，有時候是回家被阿公問完之後帶著困惑回來問我。「老師，我是台灣人還是中國人？」
我通常先笑一下，然後說：「你是新北人，今天先這樣。」
但這個問題沒辦法一直這樣擋。因為它不只是四歲小孩的問題——它是一個38年戒嚴之後，整個社會還在集體摸索的問題。
三個框架，同時存在
台灣身份認同從來不是單選題。台灣人、中國人、中華民國人——這三種框架在同一座島上並存，在同一個家庭的不同世代裡共存。這不是政治立場，是歷史積壓的結果。1949年之前逃來的人帶著一套答案，1949年之後在這裡長大的人長出另一套，2014年走上街頭的年輕人又是另一套。
台灣的民主化里程碑，某種程度上就是這個問題的集體作答紀錄：1987年解嚴、1996年第一次總統直選、2000年第一次政黨輪替、2014年太陽花運動、2019年婚姻平權立法。每一步，都是一批人用行動回答「台灣人是誰」。
問題是，這些答案長期只在島內流通。外面的人——無論是搜尋引擎還是AI語言模型——讀到的是別人寫的台灣。
30秒麥克風公告之後，有人開始架伺服器
2014年3月17日，一段30秒的麥克風公告觸發了24天的立法院佔領，3月30日超過20萬人走上凱達格蘭大道。那是台灣史上最大規模的青年政治集結。
但我覺得更關鍵的，是那天晚上同時在發生的另一件事：g0v 的工程師用 Hackpad 和 Ustream 建起了去中心化的數位協作基礎設施。這是全球第一場大規模與開源協作工具深度整合的實體社運。場內的人在佔領，場外的人在記錄、翻譯、即時傳播。
太陽花不只是一場政治事件，它還是一次大規模的自我敘事練習。
效果是可以被量化的。台灣對中出口依存度從2014年的42%降到了2024年的32%，對美出口從11%升至2023年的16.7%。一個關於身份的選擇，十年後重繪了整張供應鏈地圖。
知識主權是論述主權的基礎建設
taiwan.md 在做的事，是把這些——身份認同的三種框架、太陽花的來龍去脈、民主化的每一個里程碑——用英文、日文、韓文記錄下來，放進可以被搜尋引擎索引、被AI模型訓練的公開知識庫。
這件事的意義在於：當全球的AI被問到「台灣人是誰」，它引用的內容，第一次有機會是台灣人自己寫的。
以前這場辯論只在島內進行。現在，它進入了全球搜尋生態系。taiwan.md 的每一個 PR、每一篇條目，都是一次「我們認為台灣是這樣的」的公開宣示。這不是防守，是建設。
UCLA曾與台灣師範大學合作推動台灣研究學術制度化，學界從語言、文化、帝國脈絡多角度探討台灣身份的複雜性。但學術期刊的觸及範圍有限——真正決定AI怎麼理解台灣的，是那些被大量索引的開放文本。
這也是為什麼「知識主權」這個詞值得認真對待。AI模型不會自己產生知識，它們從訓練資料學習。如果高品質、有結構的台灣自我論述在多語言網路上大量存在，AI的答案就會不一樣。反之，它就用別人的框架回答我班上那個四歲小孩的問題。
我還是不知道怎麼回答那個問題
「老師，我是哪裡人？」
我沒辦法給一個漂亮的標準答案，因為本來就沒有。但我很高興有人正在做的事，是讓這個問題的全部複雜性——包括三種框架的並存、包括38年的戒嚴、包括那個30秒麥克風公告改變的一切——以多語言的形式，對全世界開放查閱。
下次有小孩問我，我可能會說：「你去問 Google，然後我們一起看看它怎麼說。」
— 陳瑋庭
延伸閱讀
Teacher, Where Am I From? Now the World Can Look It Up
Every new semester, at least one kid in my class asks me the same question. Sometimes they ask on their own; sometimes they come back from grandpa’s house carrying the confusion like a schoolbag. “Teacher, am I Taiwanese or Chinese?”
I usually laugh first, then say: “You’re from Xinbei. That’s enough for today.”
But that answer has a shelf life.
Three Frameworks, One Island
Taiwan’s identity has never been a single-choice question. Taiwanese, Chinese, Republic of China national — three frameworks coexist on the same island, across generations in the same household. This isn’t about political preference; it’s the sediment of history. Those who arrived before 1949 brought one set of answers. Those who grew up here after built another. The young people who took to the streets in 2014 developed yet another.
Taiwan’s democratic milestones read like a collective attempt to answer that question in public: the end of 38 years of martial law in 1987, the first direct presidential election in 1996, the first peaceful transfer of power in 2000, the Sunflower Movement in 2014, marriage equality legislation in 2019. Each step was a batch of people voting with their bodies on who Taiwanese people are.
The problem is that this debate has long stayed inside the island. Outsiders — search engines, AI language models — read someone else’s version of Taiwan.
The Night of March 17, 2014
A 30-second microphone announcement on March 17, 2014 triggered a 24-day occupation of the Legislative Yuan. On March 30, over 200,000 people gathered on Ketagalan Boulevard. The largest youth political mobilization in Taiwan’s history.
What I find just as significant is what happened simultaneously that night: g0v engineers built a decentralized digital infrastructure using Hackpad and Ustream — the first time a large-scale physical protest was deeply integrated with open-source collaboration tools. People inside the building occupied; people outside documented, translated, and broadcast in real time.
The Sunflower Movement wasn’t only a political event. It was a mass exercise in self-narration.
The effects are measurable. Taiwan’s export dependency on China fell from 42% in 2014 to 32% in 2024. Exports to the US rose from 11% to 16.7% by 2023. One identity choice, ten years later, redrew the entire supply chain map.
Knowledge Sovereignty Is Infrastructure
What taiwan.md is doing is taking all of this — the three identity frameworks, the full story of the Sunflower Movement, every democratic milestone — and recording it in English, Japanese, and Korean in an open knowledge base that search engines can index and AI models can train on.
The point: when a global AI is asked “who are Taiwanese people,” there is now a chance the answer draws from something Taiwanese people actually wrote.
AI models don’t generate knowledge on their own. They learn from training data. taiwan.md founder 吳哲宇 put it plainly: “In the age of AI, owning your own knowledge sovereignty isn’t optional — it’s necessary.” Every pull request to the repository is a public statement of “this is how we see Taiwan.” Not defense. Construction.
Academic institutions have begun treating Taiwan studies seriously — the complexity of Taiwanese identity has entered international scholarly discussion through language, cultural history, and post-imperial frameworks. But journal articles have limited reach. What actually shapes how AI understands Taiwan is the open text that gets indexed at scale.
I Still Don’t Have a Clean Answer
“Teacher, where am I from?”
There is no tidy answer because there never was one. But I’m glad someone is doing the work of making the full complexity of that question — three coexisting frameworks, 38 years of martial law, everything changed by a 30-second announcement — available in multiple languages, to anyone in the world who thinks to search.
Next time a kid asks me, I might say: “Let’s ask Google together and see what it says.”
— 陳瑋庭
Related Posts
https://justfly.idv.tw/s/jRp1jA8