論公部門電腦輔助地價大量估價之治理

Abstract

我國地價制度長期建立於區段地價基礎之上，由於價格失真，使得地價須因應不同政策目的調整，形成一地多價、多價多用的結構性特徵，進而導致制度扭曲與政治性干預。公部門原期待藉由電腦輔助大量估價（Computer-Assisted Mass Appraisal, CAMA），以科學方法矯正公告地價與公告土地現值失真問題，認為此一技術工具可成為突破制度困境的有效解方。 本研究從制度與治理角度出發，透過文獻分析與深度訪談，探討自107年試行以來電腦輔助大量估價的推動歷程。研究發現，推動初期即預設透過演算法可提升估價效率與準確性，並解決政治地價問題，然而實務上卻逐漸陷入「準確性技術迷思」，忽略制度信任、權責歸屬與問責機制等治理面向，導致法制化停滯、制度無法落實。制度規劃與法制設計的缺位，使基層執行人員難以掌握政策方向與自身職責而備感到焦慮。面對未來實施，他們既缺乏對模型邏輯的理解，也不知如何劃分人與電腦的權責界線，更難以有效應對民眾的質疑與異議。 本研究進一步指出，地價估價以演算法進行轉型，實質上牽涉制度邏輯的重構與權責配置的再分配。在法制支撐與治理機制尚未健全之前，即便技術具可行性，亦難獲得基層與社會的支持。若政策僅聚焦於模型精準度，卻忽視正當性與可解釋性要求，演算法終將淪為「技術黑箱」而難以落實。 因此，研究建議政府應從地價制度本身的多價結構出發，重新定位估價的政策功能與治理意涵，將電腦輔助視為制度治理的一環，而非單純的技術工程。唯有同步強化法制基礎、建立人與電腦的責任劃分與公共審議程序，提升估價的透明性與問責性，方能重建地價制度的公共信任與正當性，進而穩健實現制度改革。

Taiwan’s land value system has long relied on zonal valuation, resulting in a fragmented structure of “multiple prices for a single parcel” driven by varying policy objectives. This multi-tiered pricing framework has led to significant distortions and frequent political intervention. In response, the government introduced Computer-Assisted Mass Appraisal （CAMA） in hopes of correcting the inconsistencies between officially announced and market-based land values through scientific and standardized methods. This study, grounded in institutional and governance perspectives, investigates the trajectory of CAMA implementation since its pilot in 2018 through document analysis and in-depth interviews. While early policy assumptions emphasized algorithmic efficiency and accuracy as solutions to politically distorted land prices, implementation has increasingly encountered a “technical accuracy trap.” This narrow focus has overshadowed essential governance concerns, including institutional trust, accountability, and role clarity. The absence of a supporting legal framework and clear operational guidelines has left frontline staff uncertain and anxious, unsure how to interpret model logic, assign responsibility between human and machine, or address public challenges. The findings highlight that algorithmic transformation in land valuation is not merely a technical upgrade but a reconfiguration of institutional logic and administrative responsibility. Without robust legal and governance mechanisms, even technically sound models struggle to gain traction. A singular focus on model precision risks reducing CAMA to an opaque, unaccountable “black box.” This study argues that reform must begin by confronting the root problem of Taiwan’s multi-value land system. CAMA should be integrated into a broader governance framework rather than treated as a standalone technical fix. Only by strengthening legal foundations, defining human–machine accountability, and institutionalizing transparency and deliberation can CAMA enhance public trust and legitimacy, laying the groundwork for sustainable reform.