論人工智慧於僱傭關係之展望與挑戰

A Study on the Prospects and Challenges of Artificial Intelligence in Employment Relationships

隨著人工智慧技術的廣泛應用，存在一潛在風險，即雇主在利用這些技術作為招聘決策的依據時，可能蒐集和處理與招聘不直接相關的個人資訊。此行為有可能導致基於宗教、性別、性傾向等個體特徵的歧視行為。雖然我國《就業服務法》、《性別工作平等法》及其他相關法令已明文禁止此類型的歧視，但在當前科技快速發展的背景下，這些法令的實際應用仍面臨諸多挑戰。例如，亞馬遜公司曾採用一套人工智慧招聘系統，該系統因訓練數據偏向男性而使女性應徵者得分受到貶低，這一事例揭示人工智慧可能延續甚至放大社會現存的偏見。為確保人工智慧技術在僱傭過程中的公正性，並避免潛在的歧視問題，本研究建議提高演算法的透明性和可解釋性，成立專業的第三方機構進行演算法審查和評估，以及對現行的法令進行修訂等建議，以更為周全地應對人工智慧可能帶來的招聘歧視問題。

With the widespread adoption of artificial intelligence (AI) technologies, there is a latent risk that employers, when utilizing these technologies for recruitment decisions, might gather and process personal information not directly pertinent to hiring. Such practices have the potential to result in discriminatory actions based on individual characteristics like religion, gender, and sexual orientation. Although the“Employment Service Act”and the“Act of Gender Equality in Employment,”and other relevant legislation in Taiwan explicitly prohibit such discrimination, the practical application of these laws faces numerous challenges in the rapidly evolving technological landscape. For instance, Amazon once employed an AI recruitment system that, due to male-biased training data, undervalued female applicants. This case vividly illustrates the potential for AI to deteriorate the existing societal biases. To ensure fairness in using AI during the recruitment process and to prevent potential discrimination, this study strongly recommends enhancing the transparency and interpretability of algorithms. It also suggests the establishment of specialized third-party institutions for algorithmic review and assessment. Furthermore, amending existing laws is advised to more comprehensively address potential discriminatory issues arising from AI in recruitment processes.