Abstract:

The paper delves into the ethical dimensions of Artificial Intelligence in Education (AIED), focusing on the evolution of educational interactions and the advent of EdTech, highlighting the emergence of a colonialism-like dominance by major corporations in the AIED space. It starts with critical examination of the surge in venture capital investments in AIED startups, juxtaposing the economic boom against the actual utility and ethical concerns in deploying such technologies in education.

The introductory section brings to light the discrepancies between the marketed promise of AIED solutions and their real-world applicability, especially in well-resourced educational settings, drawing attention to the industry's prioritization of profit over evidence-based educational enhancements.

Central to the paper is the introduction of the "Personal Primer" (PP) project, a countermeasure to the prevailing AIED colonialism and accountability issues. Inspired by Neal Stephenson's "Diamond Age", PP is built on principles of open-source, do-it-yourself ethos. It aims to create a book-like AIED artifact that assists in literacy and fosters literacy among users, enabling them to engage with, repair, and enhance the technology. The paper underscores the project's educational and ethical grounding, detailing the development of hardware and software components that embody the project's dual objectives.

The ethical framework guiding the PP project is articulated through imperative statements that address AIED developers, emphasizing a moral and pedagogical foundation for technology development. These imperatives are rooted in Kant's categorical imperative, advocating for actions that can universally apply as moral laws.

Specifically, the paper stresses the Parental Responsibility Imperative, which mandates the creation of AIED systems that developers would allow their own children to use, reflecting a commitment to safety, appropriateness, and educational value.

This imperative is complemented by the Alignment Imperative, which calls for the development of AIED systems in congruence with the developer's moral values, ensuring that the technology aligns with what they consider suitable for their own children. A concrete example of such "alignment" is given whereby answers to questions "Can X be considered a good role model for children ?" as provided by dominant large-language models (GPT3.5, GPT4, Claude, Mistral) are compared to answers provided by real human teachers and parents.

Furthermore, the Child-Centric Imperative is introduced, emphasizing the adaptation of technology to meet the needs and capabilities of child learners, rather than forcing children to adjust to the technology. This principle is particularly relevant in the context of automatic speech recognition (ASR) systems within the PP project, where the technology is tailored to the individual linguistic nuances of child users, thereby preserving language diversity and preventing the homogenization of linguistic behaviors.

In summary, the paper presents a critical examination of the ethical challenges in the AIED industry, proposing the Primer project as a viable, ethically grounded alternative. Through a detailed discussion of the project's ethical foundations, the paper contributes to the discourse on responsible AIED development and deployment, advocating for technologies that are inclusive, equitable, and aligned with the moral and pedagogical values of their creators.

Keywords:

Artificial intelligence in education, Personal Primer, engineering ethics, parental responsibility, alignment imperative, child-centric imperative.