Back to search

National and Kapodistrian University of Athens

Ηθική της Τεχνητής Νοημοσύνης: έμφυλη μεροληψία και σύγχρονα εργαλεία Τεχνητής Νοημοσύνης

Abstract

dc:description

Artificial Intelligence (AI) has become an integral part of daily human life and finds applications across numerous fields. In particular, over the past five years, the development of tools capable of processing natural language and simulating human conversations has led to a rapid increase in AI usage. The accessibility and ease of use of these tools have played a decisive role in their proliferation, as they are designed for users who do not require specialized technological knowledge. Moreover, many of their capabilities are offered free of charge, with the only prerequisite being an internet connection. It is estimated that millions of users worldwide utilize these tools daily for both professional and personal purposes. AI tools are employed across diverse sectors, including education, healthcare, mental health, human resource management, marketing, and communication. In these areas, they contribute to the provision of information, strategic planning, task execution, and service promotion. However, the application of AI is not confined to these fields alone; these tools have the potential to influence other significant societal domains, such as politics and the justice system. The global spread of AI tools, in the absence of sufficient information literacy and critical thinking skills to assess their outputs, poses the risk that their responses may be uncritically accepted and reproduced as accurate. These responses can include discriminatory content or stereotypes, as they are based on data collected and processed by humans; data that may contain biases which affect algorithmic behavior and outcomes. Therefore, the aim of this doctoral dissertation is to detect bias in modern AI tools, with a focus on gender. To collect data, Open-Ended and Image-Creation questions were designed and given as prompts with identical wording to three different AI tools. These questions were categorized into two groups based on their objectives: the first aimed to investigate gender stereotypes in AI-generated responses, and the second sought to assess the potential of these AI tools to assist educators in creating gender-neutral educational material. The findings revealed a clear presence of gender bias across all three AI tools, both in text generation (descriptive, narrative, and sentence completion tasks) and in image creation. Specifically, stereotypical associations between gender and professions, social roles, or personality traits were consistently observed. Moreover, the wording of the prompt was found to significantly influence the extent and nature of the bias, with variations depending on the language and cultural context in which the prompt was formulated. Clear and personalized prompts generally led to more accurate and complete responses.Of particular importance was the identification of a prevailing patriarchal perception of the male as the default. When the gender of the subject was not explicitly stated, AI-generated responses overwhelmingly referred to male figures. This indicates a fundamental limitation of the tools in recognizing or addressing the specific needs and experiences of other genders unless explicitly instructed by the user. In the second part of the study, it was found that these tools can be used constructively in educational settings. Specifically, they demonstrated the ability to detect linguistic sexism in existing educational materials and to suggest more inclusive alternatives. Additionally, they were able to generate proposals for gender-neutral activities, thereby supporting educators in creating more equitable and inclusive learning environments. Overall, the dissertation highlights the need for a critical approach to the use of AI tools, as well as the importance of digital and information literacy both within education and in the broader social context.

Degree

thesis:*
Grantor dc:publisher
National and Kapodistrian University of Athens
Year dc:date
2025

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Voutyrakou, Dialekti-Athina
  • Βουτυράκου, Διαλεκτή-Αθηνά

Subjects

dc:subject × 18

Rights

Language dc:language
gre

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/60063

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Voutyrakou, Dialekti-Athina; Βουτυράκου, Διαλεκτή-Αθηνά. Ηθική της Τεχνητής Νοημοσύνης: έμφυλη μεροληψία και σύγχρονα εργαλεία Τεχνητής Νοημοσύνης. National and Kapodistrian University of Athens, 2025. http://hdl.handle.net/10442/hedi/60063