我們傾向相信數據， 因為它看起來客觀。 它記錄我們做過什麼， 而不是我們聲稱什麼。 點擊。 搜尋。 購買。 停留。 這些痕跡被視為「真實的我們」。 系統從這些記錄中學習。
辨識模式，預測偏好， 並呈現看似貼合的選項。 背後有一個簡單的假設： 如果數據反映行為， 那麼行為就反映自我。 但這個假設依賴一個前提—— 記錄必須保持完整。 如果它沒有呢？ 大數據的影響力， 來自累積，而不是單一行為。 模式在重複中形成。 穩定在時間中建立。 過去成為決定未來呈現方式的依據。 然而，當數據被篩選、重新加權、 或被悄悄調整順序時， 過去本身便開始改變。 沒有被刪除。 沒有被捏造。 只是重點轉移。 而重點，本身就足以改變方向。 推薦系統不需要創造新的慾望。 它只需要提高某些痕跡的可見度。 出現頻繁的，變得重要。 出現較少的，變得次要。 逐漸消失的，慢慢不再被考慮。 選擇仍然存在。 但方向發生改變。 選項沒有被拿走。 只是被重新排列。 而重新排列， 往往比直接限制更有效。 久而久之， 系統呈現出一個基於「調整後過去」的自我版本。 我們開始在其中認出自己。 「這很像我。」 「這很合理。」 「這是我一向會選的。」 但這種熟悉感， 未必來自中立的映照。 它可能是一個被穩定下來的敘事—— 不一致的部分被削減， 連續性被強化。 當推薦愈來愈貼合， 偏離反而變得不自然。 甚至令人不安。 選擇陌生選項， 會感覺錯誤—— 不是因為它真的錯， 而是因為它打破了被展示的模式。 偏好成為習慣。 習慣成為身份。 身份變得可預測。 數據操控最微妙的地方， 不在於欺騙。 而在於合理。 系統不需要說謊。 它只需要安排注意力。 當注意力被重塑， 期待也隨之改變。 而期待，會影響未來的行動。 問題不再是我們是否被控制。 而是我們是否仍能分辨—— 哪些慾望來自自身， 哪些是被逐步組裝而成。 在被調整邊界內做出的選擇， 依然感覺自願。 合理。 一致。 也因此，更難被質疑。 這並不意味系統本身具有惡意。 多數設計追求的是優化，而非誤導。 但優化總有方向。 當過去可以被調整， 未來就更容易被預測。 在這樣的環境中維持自由， 不一定需要拒絕科技。 而是需要打斷。 打斷那個 替我們編寫連續性的敘事。 因為我們所認識的自我， 或許並非單純來自本質， 而是來自 哪些片段的過去 仍被允許保留。
English Version
We tend to trust data because it appears objective, recording what we have actually done rather than what we claim about ourselves, as clicks, searches, purchases, and pauses are treated as evidence of a “real self” that systems can learn from, identifying patterns, predicting preferences, and presenting options that feel highly aligned with who we are, all based on a simple assumption that if data reflects behavior then behavior reflects identity, yet this assumption depends on a critical condition that the record must remain complete and neutral, and what happens if it does not, because the power of big data lies not in isolated actions but in accumulation, where patterns emerge through repetition and stability forms over time, allowing the past to shape how the future is presented, but when data is filtered, reweighted, or subtly reordered, the past itself begins to shift, not through deletion or fabrication but through changes in emphasis, and emphasis alone can redirect outcomes, as recommendation systems do not need to create new desires but only need to increase the visibility of certain traces, making what appears frequently seem important, what appears less often seem secondary, and what gradually fades away eventually disappear from consideration, so that choices still exist but their direction changes, not because options are removed but because they are rearranged, and rearrangement can be more effective than restriction, because over time the system presents a version of the self based on an adjusted past, a version that feels familiar and convincing, leading us to recognize ourselves within it and say that it makes sense or that it matches what we usually choose, even though this sense of familiarity may not come from neutral reflection but from a stabilized narrative where inconsistencies are reduced and continuity is reinforced, and as recommendations become increasingly accurate, deviation begins to feel unnatural or even uncomfortable, so that choosing something unfamiliar can feel like making a mistake, not because it truly is but because it disrupts the pattern that has been presented, turning preferences into habits, habits into identity, and identity into something predictable, and the most subtle aspect of data influence is not deception but plausibility, as systems do not need to lie but only need to organize attention, and when attention is reshaped expectations follow, influencing future actions in ways that feel voluntary, reasonable, and consistent, making them difficult to question, and this raises a deeper issue not of whether we are being controlled but whether we can still distinguish which desires originate from ourselves and which are gradually assembled through exposure and repetition, because choices made within adjusted boundaries still feel like our own, and this is precisely why they are powerful, and while this does not imply malicious intent, as most systems are designed to optimize rather than mislead, optimization always follows a direction, and when the past can be adjusted the future becomes easier to predict, so maintaining freedom in such an environment does not necessarily require rejecting technology but requires interruption, moments that break the continuous narrative being constructed for us, because the self we recognize may not arise purely from essence but from which fragments of the past are allowed to remain visible, and in questioning that visibility we begin to see that our choices may be less fixed than they appear, and that within even a subtly shaped system there still exists the possibility of stepping outside the pattern, if only we are willing to notice it.