This document has been created as a result of the collaboration between the participants and the advisors who attended the “Artificial Intelligence for Social Well-being Summit” organized by the Robert College Social Entrepreneurship, Innovation and Creativity Office on May 21, 2021, hosting 60 students from 13 schools. This document and the summit were inspired by the Montréal Declaration for a Responsible Development of Artificial Intelligence.
AI Guiding Principles
Purpose: Artificial intelligence is estimated to possess the potential to radically transform the lives of individuals and institutions. Design and development processes and practices will also figure in the near future as a prevalent technological factor at schools. As with other technological applications, it is essential to define the fundamental principles for the development of artificial intelligence applications. This becomes especially paramount when it is a question of artificial intelligence drawing on the personal datasets of individuals. This document has been compiled to serve as a fundamental principles guide to the schools, which are institutions that play a leadership role in investing and using the artificial intelligence technologies. Each school may adopt these principles as they are or refer to them to develop its own fundamental principles.
Intended Audience: This document has been written for the schools that plan to design, develop, or use artificial intelligence technologies.
How to Read this Document: A school that adopts the fundamental principles recommended in this document makes sure that the design, development or use of artificial intelligence technologies are not in contradiction with these principles. When a school decides to design, develop, or use an AI application, the constituents who are involved (students, teachers, staff) assume that the school ensured compatibility with the accepted principles. Accordingly, it is expected that the school will not put into practice any application before checking its congruity. In addition, the school will regularly check that AI applications in use are not contradictory to the adopted principles.
How to Use this Document: A school may accept these principles verbatim or modify them to suit its own dynamics. This decision will be taken at the conclusion of an evaluation process with the participation of the student, teacher, and school staff representatives. The ultimate document will be submitted to the approval of the school administration. The administration will know that these will become standing principles valid for all future AI applications and will assess them accordingly before approving. The school regularly reviews the accepted document and makes necessary changes if need be. The approved document is shared with all the constituents of the school.
The following principles are adhered to during the design and the implementation stages of the AI technologies in schools.
- Privacy Principle: The privacy of the individuals is respected.
- Awareness Principle: The individuals involved in the development and use of AI are informed, and their consent be requested.
- Inclusion Principle: The diversity of the data sets and the needs of the individuals’ concerns are respected during the design and the use of AI.
- Personalization Principle: The AI practices offer a personalized learning and life experience to the students.
- Transparency Principle: The method of use and the results of the data collected for the development and use of AI are freely available to the data owners.
- Objectivity Principle: AI is expected to contribute to generating objective results without any bias.
- Well-being Principle: AI contributes to the psychological and physical well-being of the individuals.
- Sustainability Principle: The AI designs and applications contribute to the environment and to its sustainable development.
1. On the Privacy Principle: Since AI applications are also based on personal datasets, care is taken not to reveal the identity and the privacy of the individual when it is necessary to use their personal data, and these datasets must be anonymized. If the AI applications concern the growth of an individual (e.g., individualized learning apps), the other individuals are denied access to the dataset.
2. On the Awareness Principle: Information is given to the owners of the bulk datasets to be used in AI apps, and they are asked to give consent when needed. In addition, training sessions are offered to students, parents, teachers, and staff about the nature of the AI technology, the correct use of such apps, etc. In that respect, content can be integrated into the curriculum, electives can be offered, workshops can be organized, seminars and conferences can be held.
3. On the Inclusion Principle: Inclusion means to make sure that individuals do not feel excluded because of their differences. To ensure this when designing or developing AI applications, the datasets used to feed the respective algorithms are created heterogeneously. For example, if a face recognition AI is used, it is confirmed that it is sensitive to personal traits like skin color, gender, etc. The AI apps also help strengthen the diversity and the solidarity within the institution.
4. On the Individualization Principle: The AI apps offer individualized instruction and experiences to the students. The schools use their own student datasets if possible. Schools that do not have records of past data may use the AI apps developed with general datasets. With time, these schools should reflect their own students’ data onto the AI and localize it. Examples of individualized learning: deciding the content of projects/homework, choosing the electives, improving the living conditions at the school.
5. On the Transparency Principle: The data owners should be given clear and comparative information about how and to what purpose their datasets will be used, and the data which have been used should be shared with them in case they so request.
6. On the Objectivity Principle: One of the major issues related to AI in the current climate is bias. The more heterogeneous the feed datasets are, the more objective is the algorithm output. The AI algorithms are not biased per se. Therefore, it must be remembered that any bias displayed by the AI will originate from the school’s mode of operation in that area. Even when the datasets used for the design or the development of an AI app are the past data of the school, the bias dimension must be analyzed, and if any bias is in question, this must be tackled as a problem to be solved.
7. On the Well-being Principle: The AI apps improve the individuals’ living conditions, health, and working conditions. AI does not prevent the individuals from using their intellectual and physical abilities to full extent. The digital domain created at the school does not increase the individuals’ stress or concerns, it does not make them feel as if others had intruded into their private sphere.
8. On the Sustainability Principle: The AI designs or applications developed within the school should not jeopardize the ecological systems. Applications which contribute to environmental sustainability should be preferred. The AI infrastructure aims for the efficient use of energy.