人工智慧中的其中一個重要技術—自然語言處理，透過理解及生成人類自 然語言，如ChatGPT的聊天機器人程式等更是逐漸滲透我們的工作和生活。但 不同研究指出自然語言處理裡隱含了性別偏見，揭示了因人工智慧複製了人類 社會的性別偏見而將偏見重演在其生成文字上。文字對人類文明文化有著舉足 輕重的影響，若這種帶有偏見的生成文字毫無覺察地出現在我們日常中，又將 會產生甚麼問題呢?本研究欲探討自然語言處理的不同面向，以及它如何改變 性別平等的現狀，並以推測設計的思維去推測—若我們放任這問題，當二十年 後人工智慧更成熟及全面地發展後，將如何影響我們的性別觀念，最後以策展 設計呈現自然語言處理對性別偏見的影響，將資料及概念傳遞給公眾，以喚起 公眾關注。
從文獻探討中，本研究首先探討了自然語言處理的基本釋義及其不同應用 範圍。之後，則是研究自然語言處理技術帶來的性別偏見問題。性別偏見包括 誹謗、刻板印象、識別偏差和代表性不足。這些偏見在人工智慧系統中被放 大，可能導致性別不平等的惡性循環。故此，再在文獻探討中探討推測設計的 設計思維及設計方法，從界定「假如...」問題、情境設定等方法去為後續的研 究創作奠定基石。最後，透過研究策展設計的定義及策展界入社會議題的相關 論述，希望透過展覽激發公眾更多的討論和關注。
在創作過程中，透過推測設計的思維及設計方法，與及現今正愈趨頻繁使 用自然語言處理技術的範疇，最後聚焦在人類使用文字的三大重要領域:歷 史、新聞及法律。本研究創作了一個名為《未來性別:對話》的展覽，分為兩 大區，「現在區」以三個展品描述自然語言處理的資訊，及其產生性別偏見的 例子、原因及面貌。而「未來區」則以三個展品:「未來編年史」、「未來新 聞生成器」及「未來法庭」引發觀者對未來的想像及激發關注討論。
本研究於 2024 年 7 月 4 日至 7 月 8 日在國立台灣科技大學·大用工坊舉 辦《未來性別:對話》的展覽，共展出 7 個設計作品。
關鍵字:自然語言處理技術、性別偏見、推測設計、策展設計

One of the key technologies in AI is natural language processing (NLP), which enables understanding and generating human natural language, as exemplified by chatbot programs like ChatGPT, which are increasingly integrated into our work and life. However, research has shown that NLP inherently contains gender biases, revealing that AI systems reproduce the gender biases present in human society through the text they generate. Language has a profound influence on human civilization and culture. If biased text generation goes unnoticed daily, what problems might it cause?
This research aims to explore different aspects of NLP and how it is changing the status quo of gender equality. Using a speculative design mindset, it speculates on how gender perceptions may be impacted when AI becomes more mature and widespread in 20 years. Finally, the research presents the findings through a curatorial design that conveys the impact of NLP on gender bias to the public, aiming to raise awareness.
The literature review first discusses the basic definition of NLP and its various applications. It then examines the gender bias issues brought about by NLP, including defamation, stereotypes, identification biases, and underrepresentation. These biases are amplified in AI systems, potentially leading to a vicious cycle of gender inequality. The research then explores the design thinking and speculative design methods, such as defining "what if" questions and setting scenarios, to lay the foundation for the subsequent creative work. Lastly, through the lens of curatorial design and its relevance to addressing social issues, the research hopes to spark more public discussion and attention through the exhibition.
In the creative process, guided by speculative design and focusing on domains where natural language processing is increasingly prevalent - history, news, and law - the research created an exhibition titled "Future Gender: Dialogue". It is divided into two main sections: the "Present" section showcases three exhibits that describe NLP, its gender bias examples, causes, and manifestations; the "Future" section presents three exhibits: "Future Chronicle", "Future News Generator", and "Future Courtroom", to evoke the audience's imagination and stimulate discussion about the future. The "Future Gender Dialogues" exhibition was held at the National Taiwan University of Science and Technology's Design Lab from July 4 to July 8, 2024, showcasing 7 design works.
Keywords: Natural Language Processing, Gender Bias, Speculative Design, Curatorial Design

外文文獻
1. Auger,J.(2013)."Speculativedesign:craftingthespeculation".Digital Creativity, Vol. 24, pp. 11–35.
2. Banea,C.,Mihalcea,R.,Wiebe,J.,andHassan,S.(2008).Multilingual Subjectivity Analysis Using Machine Translation. Proceedings of the Association for Computational Linguistics 12th Annual Conference on Empirical Methods in Natural Language Processing (EMNLP), Honolulu, October 25-27.
3. Bolukbasi,T.,Chang,K.W.,Zou,J.,Saligrama,V.andKalai,A.(2016). Man Is to Computer Programmer As Woman Is to Homemaker? Debiasing Word Embeddings. In Neural Information Processing Systems (NIPS'16).
4. Burns,K.,Hendricks,L.A.,Darrell,T.,Rohrbach,A.andSaenko,K.(2018). Women Also Snowboard: Overcoming Bias in Captioning Models. European Conference on Computer Vision (EECV'18).
5. Chowdhary,K.R.(2020).Naturallanguageprocessing.Fundamentalsof artificial intelligence, 603-649.
6. Chowdhury,K.R.andMulani,N.(2018).AuditingAlgorithmsforBias. Harvard Business Review, https://hbr.org/2018/10/ auditing-algorithms-for- bias
7. Cohen,R.(2008).ArtificialBiologicalClock.Retrievedfromhttps:// www.cohenvanbalen.com/work/artificial-biological-clock
8. Collett,C.andDillon,S.(2019).AIandGenderFourProposalsforFuture Research.
9. Crawford,K.(2017).TheTroubleWithBias.KeynoteatNeuralInformation Processing Systems (NIPS'17).
10. Dale, R. (2011), Automated Writing Assistance. Centre for Language Technology, Macquarie University. Retrieved from https:// www.slideshare.net/slideshow/tarragona-summer-schoolautomated-writing- assistance-topic-5/9381644#8
11. Douglas, L. 2017. AI is not Just Learning our Bi- ases; It Is Amplifying Them. https://bit.ly/ 2zRvGhH. Accessed on 11.15.2018.
12. Eckes, T. And Trautner, H. M. (2000). Developmental social psychology of gender: An integrative framework.
13. George, A. (2015). The Curator's Handbook: Museums, Commercial Galleries, Independent Spaces.
14. Ghosh, S. (2021). Towards Data Science. Retrieved from https:// towardsdatascience.com/the-forceful-and-insistent-application-of-gender- stereotypes-by-google-translate-d6b79cbee48
15. Horváth, I. (2007). Comparison of three methodological approaches of design research, International Conference on Engineering Design, ICED'07, 28-31
August 2007.
16. Jamnik, M. (2005). Computer Scientist and a Woman? University of Cambridge.
17. Johannessen, L., Keitsch, M. and Pettersen, I. N. (2019). Speculative and Critical Design — Features， Methods， and Practices. Proceedings of the Design Society: International Conference on Engineering Design. 1 (1): 1623–1632.
18. Kalyanathaya, K. P., Akila, D., and Rajesh, P. (2019). Advances in natural language processing–a survey of current research trends, development tools and industry applications. International Journal of Recent Technology and Engineering, 7(5C), 199-202.
19. Koskinen, I., Zimmerman, J., Binder, T., Redström, J. and Wensveen, S. (2011). Design Research Through Practice: From the Lab, Field, and
Showroom.
20. Koskinen, I., Zimmerman, J., Binder, T., Redström, J. and Wensveen, S. (2012). Constructive Design Research.
21. Lambrecht, A., and Tucker, C. E. (2016). Algorithmic bias? An empirical study into apparent gender-based discrimination in the display of STEM career ads. SSRN Journal. Retrieved from http://www.ssrn.com/ abstract=2852260
22. Lind, M. (2021). Situating the Curatorial. Retrieved from https://www.e- flux.com/journal/116/378689/situating-the-curatorial/
23. Lu, K., Mardziel, P., Wu, F., Amancharla, P. and Datta, A. (2018). Gender Bias in Neural Natural Language Processing.
24. Mantha, Y. and Hudson, S. (2018). Estimating the Gender Ratio of AI Researchers Around the World. [online] Element AI Lab. Retrieved from https://medium.com/element-ai-research-lab/estimating-the-gender-ratio-of-ai- researchers-around-the-world-81d2b8dbe9c3
25. Minkov, E., Toutanova, K., and Suzuki, H. (2007). Generating Complex Morphology for Machine Translation. Proceedings of the 45th Annual Meeting of the Association of Computational Linguistics, June 23-30, Prague.
26. Mitrović, I. (2015), An Introduction to Speculative Design Practice. Split: Department for Visual Communications Design, Arts Academy, University
of Split.
27. Park, J. H., Shin, J. and Fung, P. (2018). Reducing Gender Bias in Abusive Language Detection. In Empirical Methods of Natural Language Processing (EMNLP'18).
28. Portal, F. (2012). Facestate: Metahaven. Retrieved from https:// www.researchgate.net/publication/353644519_Facestate_Metahaven
29. Reilly, M. (2018). Curatorial Activism: Towards an Ethics of Curating. London: Thames and Hudson.
30. Rothchild, J. (2014). Gender Bias.
31. Snyder, K. (2023). Textio. Retrieved from https://textio.com/blog/chatgpt- writes-performance-feedback/99766000464? utm_campaign=hrb&utm_medium=newsletter&utm_source=morning_brew
32. Sputniko! (2010). Menstruation Machine. Retrieved from https:// sputniko.com/Menstruation-Machine
33. Sputniko! (2010). Sushiborg Yukari. Retrieved from https://sputniko.com/ Sushiborg-Yukari
34. Sun, T., Gaut, A., Tang, S., Huang, Y., ElSherief, M., Zhao, J., Mirza, D., Belding, E., Chang, K.-W., and Wang, W. Y. (2019). Mitigating Gender Bias in Natural Language Processing: Literature Review. arXiv preprint arXiv:1906.08976.
35. Tatman, R. (2017). Gender and Dialect Bias in YouTube's Automatic Captions. In Proceedings of the First ACL Workshop on Ethics in Natural Language Processing (ACL'17), pages 53–59.
36. UNESCO. (2024). Systematic Prejudices An Investigation into Bias Against Women and Girls in Large Language Models.
37. UK Intellectual Property Office Informatics Team. (2016). Gender Profiles in Worldwide Patenting: An analysis of female inventorship.
38. Wakefield, J. (2023). Is AI about to transform the legal profession? Retrieved from https://www.bbc.com/news/business-67121212
39. Whatever.co, (2016). 79% Work Clock. Retrieved from https://whatever.co/ zh-hant/work/79-work-clock/
40. World Intellectual Property Organization. (2019). WIPO Technology Trends 2019: 113 Artificial Intelligence. ( ISBN : 978-92-805-3007-0 ) . Retrieved from https://www.wipo.int/edocs/pubdocs/en/wipo_pub_1055.pdf
41. XYXLAB Gender + Place. Hypersext City. (2021). Retrieved from https:// hypersextcity.com/
42. Zhao, J., Wang, T., Yatskar, M., Or- donez, V. and Chang, K. W. (2017). Men Also Like Shopping: Reducing Gender Bias Amplification Us- ing Corpus- Level Constraints. In Empirical Methods of Natural Language Processing (EMNLP'17).
43. Zhao, J., Wang, T., Yatskar, M., Or- donez, V. and Chang, K. W.(2018). Gender Bias in Coreference Resolution: Evaluation and Debiasing Methods. In North American Chapter of the Association for Computational Linguistics (NAACL'18).
44. Zhao, J., Zhou, Y., Li, Z., Wang, W., and Chang, K. W. (2018). Learning Gender-Neutral Word Embeddings. arXiv preprint arXiv:1809.01496.

中文文獻
1. AnthonyDunneandFionaRaby，洪世民譯(2019)，《推測設計——設 計、想象與社會夢想》。何樵暐工作室出版。
2. Chang,L.(2018)。「男人國」裡的程式女孩──矽谷的性別鴻溝有多嚴 重，又該如何翻轉?檢自 https://crossing.cw.com.tw/article/10567
3. 人文與社會科學簡訊(2019)。 21卷1期‧108年12月。檢自 https://www.nstc.gov.tw/nstc/attachments/bba9c17d-76d1-4f30-a5a1- b7204f192b96
4. 王俊傑(2004)。猶如獨立製片的展覽策劃—關於當代展覽的機制與實 踐經驗。 《國際交流的拋物線連結—2004 策展人論壇文集》，國立台彎 美術館，168-174。
5. 呂佩怡(2013)。策展(Curating)/策展(Curation)?。檢自https:// mag.ncafroc.org.tw/article_detail.html? id=297ef722719c827a0171aad1b4b6001d
6. 呂純如(2005)。七年級社會學習領域性別平權教育之研究-教科書分析 與教室觀察。 國立臺灣師範大學公民教育與活動領導學系碩士論文，未 出版，臺北市。
7. 賴禹安(2022)。人工智慧大躍進，但AI會取代新聞工作者嗎?檢自 https://feja.org.tw/67719
8. 李宏毅(2020)。斷開中文的鎖鍊!自然語言處理(NLP)是什麼?檢自 https://aiacademy.tw/what-is-nlp-natural-language-processing/
9. 周素碧(2010)。國中國文教科書性別意識型態之內容分析-以九十七學 年度南一、康軒、翰林三版本為例。國立中正大學教育學研究所碩士論 文，未出版，嘉義縣。
10. 林宏璋(2009)。策展主體。典藏今藝術，203。
11. 林慧敏(2022)。人力資本與性別對升遷結果的影響: 探討性別偏見與工 作-家庭衝突之干擾效果 ，國立臺南大學經營與管理學系科技管理碩士 在職專班碩士論文，未出版，臺南市。
12. 侯奇江(2022)。後人類離性別平等有多遠1:家暴AI女友與人機關係 的反身性。檢自https://m.thepaper.cn/newsDetail_forward_16721772
13. 胡昌明(2018)。被告身分差異對量刑的影響。《清華法學》2018年第4 期第91-110頁。
14. 馬琳編著(2014)。展覽改變藝術:策展的可能性。上海，上海大學。
15. 施懿珊(2023)。施懿珊個展:數位孿生技術下的未來判罰形式。檢自 https://punishment-20302.webnode.tw/
16. 國立臺灣科學教育館(2022)。設計我們的世界-科技性別化創新展覽。 檢自https://www.ntsec.gov.tw/article/detail.aspx?a=154
17. 陳信如(2023)。人人都能請 AI 當律師?AI 如何翻轉法律圈生態?檢 自https://sunrisemedium.com/p/318/ai-in-law
18. 陳俊良(2014)。策略策展，國立台灣藝術大學創意產業設計研究所博 士論文。
19. 郭旗成(2023)。伴侶間親密關係與性別偏見之研究。樹德科技大學。
20. 曾元顯(2012)。自然語言處理。國家教育研究院樂詞網。
21. 錢陳翔(2018)。策展人策展流程及構成要素之研究。