سوگیری الگوریتمی و عدالت آموزشی در عصر هوش مصنوعی، پیامدهای اجتماعی و راهکارهای سیاستی در ایران

نوع مقاله : مقاله پژوهشی

نویسندگان

1 دانشجوی دکتری مهندسی فناوری اطلاعات، گروه مهندسی کامپیوتر و فناوری اطلاعات ، دانشگاه قم، قم، ایران.

2 استادیار گروه مهندسی کامپیوتر و فناوری اطلاعات، دانشگاه قم، قم، ایران

10.22091/jrim.2026.14502.1436

چکیده

با گسترش کاربرد هوش مصنوعی و به‌ویژه مدل‌های زبانی بزرگ در آموزش، این پرسش مطرح می‌شود که این فناوری‌ها چگونه می‌توانند عدالت آموزشی را تقویت یا تضعیف کنند. هدف این پژوهش، تبیین ابعاد سوگیری الگوریتمی در نظام‌های آموزشی و استخراج پیامدهای اجتماعی و راهکارهای سیاستی متناسب با بافت ایران است. پژوهش حاضر کیفی و از نوع تحلیل اسنادی ـ تحلیلی است و با استفاده از تحلیل محتوای نظام‌مند گزارش‌های رسمی، مقالات علمی و مطالعات موردی بین‌المللی، براساس چک‌لیست مفهومی عدالت آموزشی، انجام شده است. چارچوب نظری مطالعه بر نظریه عدالت به‌مثابه انصافِ جان راولز و رویکرد قابلیت‌های آمارتیا سن استوار است. یافته‌ها نشان می‌دهد سوگیری در چهار سطح تعریف مسئله، داده، مدل‌سازی و تفسیر/ پیاده‌سازی می‌تواند نابرابری‌های آموزشی را بازتولید کند و در بافت متنوع و دارای شکاف دیجیتال ایران این نابرابری‌ها را تعمیق بخشد. بر این اساس، شش محور سیاستی شامل طراحی عدالت‌محور الگوریتم‌ها، پایش تنوع داده‌ها، آموزش اخلاق الگوریتمی به ذی‌نفعان، شفافیت و پاسخ‌گویی، توسعه مدل‌های بومی و کاهش شکاف دیجیتال پیشنهاد می‌شود. نوآوری مقاله در پیوند دادن نظریه‌های عدالت با ادبیات سوگیری الگوریتمی و بسط چارچوبی بومی برای سیاست‌گذاری عدالت آموزشی در عصر هوش مصنوعی در ایران است.

تازه های تحقیق

براساس تحلیل اسنادی و تحلیل محتوای کیفی، این پژوهش نشان داد که سوگیری الگوریتمی در آموزش، پدیده‌ای محدود به خطای فنی یا نقص در طراحی یک مدل خاص نیست، بلکه در چند سطح به عدالت آموزشی آسیب می‌زند. مضامین استخراج‌شده در بخش یافته‌ها – شامل بازتولید نابرابری‌های ساختاری، تعارض میان دقت الگوریتمی و عدالت آموزشی، سوگیری اتوماسیون و تفکیک میان ضررهای تخصیصی و بازنمایی – در مجموع ترسیم می‌کنند که چگونه هوش مصنوعی می‌تواند نابرابری‌های پیشین را در قالبی فناورانه بازتولید کند.

این یافته‌ها در پرتو چارچوب‌های نظری عدالت، معنا و عمق بیشتری پیدا می‌کنند.

مقایسه این چارچوب‌ها با شواهد مرور شده نشان می‌دهد که بسیاری از کاربردهای فعلی هوش مصنوعی – حتی زمانی که از نظر دقت کلی موفق‌اند – با این معیارهای عدالت هم‌راستا نیستند، زیرا خطای بیشتری برای گروه‌های حاشیه‌ای تولید می‌کنند یا دسترسی آن‌ها به فرصت‌های آموزشی را محدود می‌سازند. به بیان دیگر، «موفقیت فنی» الگوریتم‌ها لزوماً به «عدالت آموزشی» منجر نمی‌شود و بدون وارد کردن شاخص‌های عدالت توزیعی، ارزیابی صرفاً براساس دقت می‌تواند گمراه‌کننده باشد.

وجه دیگر نتایج، به سطح کنش انسانی و تفسیر الگوریتم‌ها مربوط است. شواهد مربوط به سوگیری اتوماسیون نشان داد که معلمان، مدیران و سیاست‌گذاران در مواجهه با سامانه‌های هوش مصنوعی ممکن است تصمیم‌های الگوریتمی را به‌طور پیش‌فرض معتبرتر از قضاوت انسانی تلقی کنند و در عمل نقش انتقادی خود را به «تاییدکننده خروجی سیستم» تقلیل دهند. در کنار این، مطالعات اخیر درباره مدل‌های زبانی بزرگ نشان می‌دهد که این مدل‌ها، بر اثر آموزش بر داده‌های عمدتاً غربی، گرایش به بازتولید کلیشه‌های جنسیتی و فرهنگی دارند و حضور زبان‌ها و فرهنگ‌های غیرغربی را به‌طور محدودتر بازنمایی می‌کنند. ترکیب این دو روند در زمینه‌ای مانند ایران  با تنوع زبانی و فرهنگی بالا می‌تواند هم در سطح تخصیص فرصت‌ها مانند پیشنهاد دوره‌ها، بورسیه‌ها یا سطح‌بندی تحصیلی و هم در سطح بازنمایی نمادین مانند تصویری که دانش‌آموزان از خود و دیگران می‌سازند پیامدهای نابرابر ایجاد کند.

کلیدواژه‌ها


عنوان مقاله [English]

Algorithmic Bias and Educational Justice in the Age of Artificial Intelligence: Social Implications and Policy Solutions in Iran

نویسندگان [English]

  • Faegheh Faghih Moussavi 1
  • Faranak Fotouhi Ghazvini 2
1 Ph.D. Student in Information Technology Engineering, Department of Computer Engineering and Information Technology, Faculty of Engineering, University of Qom, Qom, Iran
2 . Assistant Professor, Department of Computer Engineering and Information Technology, Faculty of Engineering, University of Qom, Qom, Iran
چکیده [English]

With the expanding use of artificial intelligence—especially large language models (LLMs)—in education, a key question arises: how can these technologies strengthen or undermine educational equity? This study aims to elucidate the dimensions of algorithmic bias within educational systems and to derive social implications and policy responses appropriate to the Iranian context. The present research is qualitative, adopting a documentary–analytical approach. It employs a systematic content analysis of official reports, scholarly articles, and international case studies, guided by a conceptual checklist of educational equity. The study’s theoretical framework draws on John Rawls’s theory of justice as fairness and Amartya Sen’s capability approach. The findings indicate that bias operating at four levels—problem formulation, data, modelling, and interpretation/implementation—can reproduce educational inequalities and, in Iran’s diverse context marked by a digital divide, further intensify them. Accordingly, six policy directions are proposed: equity-oriented algorithm design; monitoring and ensuring data diversity; providing algorithmic ethics education for stakeholders; strengthening transparency and accountability; developing indigenous models; and reducing the digital divide. The article’s contribution lies in linking theories of justice with the literature on algorithmic bias and in advancing a locally grounded framework for educational equity policymaking in Iran in the age of artificial intelligence.

کلیدواژه‌ها [English]

  • algorithmic bias
  • Educational justice
  • Artificial intelligence
Alon-Barkat, S., & Busuioc, M. (2023). Human–AI interactions in public sector decision making:“automation bias” and “selective adherence” to algorithmic advice. Journal of Public Administration Research and Theory, 33(1), 153-169.
An, H., Acquaye, C., Wang, C., Li, Z., & Rudinger, R. (2024). Do Large Language Models Discriminate in Hiring Decisions on the Basis of Race, Ethnicity, and Gender? arXiv preprint arXiv:2406.10486.
Angwin, J., Larson, J., Mattu, S., & Kirchner, L. (2022). Machine bias. In Ethics of data and analytics (pp. 254-264). Auerbach Publications.
Baker, R. S., & Hawn, A. (2022). Algorithmic bias in education. International Journal of Artificial Intelligence in Education, 1-41.
Boateng, O., & Boateng, B. (2025). Algorithmic bias in educational systems: Examining the impact of AI-driven decision making in modern education. World Journal of Advanced Research and Reviews, 25(1), 2012-2017.
Bulathwela, S., Pérez-Ortiz, M., Holloway, C., Cukurova, M., & Shawe-Taylor, J. (2024). Artificial intelligence alone will not democratise education: On educational inequality, techno-solutionism and inclusive tools. Sustainability, 16(2), 781.
Carragher, D. J., Sturman, D., & Hancock, P. J. (2024). Trust in automation and the accuracy of human–algorithm teams performing one-to-one face matching tasks. Cognitive Research: Principles and Implications, 9(1), 41.
Cheng, H., Guo, Y., Guo, Q., Yang, M., Gan, T., & Nie, L. (2024). Social debiasing for fair multi-modal llms. arXiv preprint arXiv:2408.06569.
Dastin, J. (2022). Amazon scraps secret AI recruiting tool that showed bias against women. In Ethics of data and analytics (pp. 296-299). Auerbach Publications.
Esmer, S. (2021). Amartya Sen‘s capability approach and its relation with John Rawls ‘justice as fairness. Middle East Technical University.
Guo, Y., Guo, M., Su, J., Yang, Z., Zhu, M., Li, H., Qiu, M., & Liu, S. S. (2024). Bias in large language models: Origin, evaluation, and mitigation. arXiv preprint arXiv:2411.10915.
Hern, A. (2020). Ofqual’s A-level algorithm: Why did it fail to make the grade. The Guardian, 21.
Holmes, W., & Miao, F. (2023). Guidance for generative AI in education and research. UNESCO Publishing.
Kizilcec, R. F., & Lee, H. (2022). Algorithmic fairness in education. In The ethics of artificial intelligence in education (pp. 174-202). Routledge.
Mallett, B. (2023). Reviewing the impact of OFQUAL’s assessment ‘algorithm’on racial inequalities. In COVID-19 and Racism (pp. 187-198). Policy Press.
Nabi, d., Shahraki, H., Ghofran Mazloom, I., & Absalan, R. (2024). Artificial Intelligence and Reducing Educational Discrimination. The First National Conference on Modern Perspectives on Educational Issues
Nabipour Gisi, E., Ahmadi, A., Darabi, J., & Sharifi, R. (2024). Artificial intelligence and educational equity: how can technology reduce inequalities? The First National Conference on New Approaches to Educational Issues, Ramshir.
Nazari, F., Pirootiaghdam, M., & Zovko, M.-E. (2022). Educational inequalities in Iran based on the viewpoints of educational experts and qualified high school teachers. Distinctio: Journal of Intersubjective Studies, 1(2), 73-93.
Rawls, J. (2017). A theory of justice. In Applied ethics (pp. 21-29). Routledge.
Sarafa, O. I., & Oyewole, S. (2023). John Rawls on the theory of justice. Classical Theorists in the Social Sciences: From Western Ideas to African Realities, 347-375.
Sen, A. (2008). The idea of justice. Journal of Human Development, 9(3), 331-342.
Tao, Y., Viberg, O., Baker, R. S., & Kizilcec, R. F. (2024). Cultural bias and cultural alignment of large language models. PNAS nexus, 3(9), pgae346.