The advent of artificial intelligence has assumed prominence amongst all industries and various facets of people's personal lives. The integration of AI in education has been inevitable, given the significance and role of information, knowledge production and administration in the sector. This is especially so as its capabilities entail replicating higher-order thinking. Besides assisting in the education process, it also brings the element of real-life relevance, allowing education to be imparted against the backdrop of the evolving world due to the same AI. It tends to have implications on the subject matter that needs to be imparted, which tends to be something that constantly needs to answer the question of "Why and how is this particular subject matter relevant for learning?".
This induces policy-makers and educational institutions to rethink what they need to impart as knowledge, the area of matter, and the manner of thinking to be emphasised. This is because education is tasked with shaping people as individuals and equipping them with the necessary skills and knowledge to function in the world. Therefore, it is to be ensured education brings forth the instructional design for making the best of AI in the coming world instead of being invalidated by it.
The UNESCO and Education
The United Nations Educational, Scientific and Cultural Organisation (UNESCO) has committed itself to supporting its member states in harnessing AI's potential. The following primary target that is sought to be achieved is the Education 2030 Agenda, which is a build-on to the Millennium Development Goals (MDGs), and the "2015 Incheon Declaration and Education 2030 Framework for Action", whose framework entails a focus on inclusiveness and equitability in the quality of education, besides promoting lifelong learning opportunities. The 2030 Agenda would be monitored by the High-Level Political Forum (HLPF) on an international level and entails seventeen Sustainable Development Goals (SDGs), among which SDG 4 concerns itself with education.
November 2021 saw the inauguration of the Global Education Cooperation Mechanism (GCM) at the global meeting convened for SDG 4. The High-Level Steering Committee (HLSC) would carry out the governance role, comprising decision-makers worldwide. The committee would be responsible for the Transforming Education Summit follow-up, as well as contributing to the educational dimension of the Summit of the Future 2023, which is expected to be held on 22-23 September 2024, as per the General Assembly Resolution A/RES/76/307.
An emphasis has also been laid on reaching the marginalised, as evinced by the "Collective Consultation of NGOs on Education 2030". The 2030 Agenda also concerns sustainable economic growth, addressing systemic and broader governance issues, and facilitating human rights. The UN also provides data on thematic indicators through its annual "Global Education Monitoring Reports".
Background
Generative AI: Generative Artificial Intelligence, known as Generative AI or GenAI, involves new content being produced, appearing in formats representative of human thinking, in response to prompts. However, its limitations lie in its inability to create new ideas and solutions and its questionable accuracy, requiring the user to possess firsthand existing, deep-rooted knowledge. It is trained using data from various online sources, including web pages, online conversations, and mixed media. Generative AI, relating to text and images, uses algorithms that form a part of machine learning. Text-generative AI such as Chat GPT uses an artificial neural network known as a generative pre-trained transformer, while image-generative music generative AI is typically known to use Generative Adversarial Networks. GenAI uses parameters that serve to fine-tune the performance of the AI. In the case of text GenAI, growing amounts of data are fed to train the parameters that keep growing in number exponentially. At the same time, image GenAI uses a generator and a discriminator. As the names suggest, the generator generates the images while the discriminator attempts to distinguish between what is created and an actual image. The result is then used to adjust the parameters, and the said process is repeated to reduce the distinction between the generated and the authentic images.
Generative AI and Education: The use of GenAI brings forth a greater level of personalisation in learning, expe adjusting by adjusting to the needs, learning styles and patterns, and preferences of individuals. Additional resources can be provided where and when needed, and educators can save time by automating the grading and feedback process. Educational GPT involves models being trained with data specifically to meet educational purposes. This entails refinement in the model that is otherwise a derivation of humongous volumes of general training data, with lesser albeit higher-quality, domain-specific data. The targeting of the co-design of the curriculum would allow for the generation of the appropriate educational materials as well as lesson plans and other interactions and activities that seek to further the goal of the pedagogy, curriculum objectives and level of difficulty.
However, while EdGPT models are theoretically supposed to contain fewer errors and biases, errors might continue to be generated unless the GenAI model and the approach that underlies it are changed significantly. Refining the foundation models for greater targeted use of GPT in education is still at an early stage. Besides adding relevant subject knowledge and removing biases, refinement will also be done by adding pertinent learning knowledge models.
Other challenges that require addressing are the extent to which the said EdGPT may extend beyond the subject matter in furtherance of the pedagogy and interactions, the extent of ethical collection of teacher and learner data, and ensuring that neither the students' human rights nor the teachers are undermined or disempowered.
AI and Education - Controversies
Worsening Digital Poverty: AI requires tremendous computing data and processing power, narrowing its availability to only a few technology companies and economies. This leads to a divide between those who do and those who need the means to create and control GenAI. The resulting "data-poor"" nations "would get excluded and possibly "colonised"" by the em"added standards in GPT models.
Outpacing National Regulatory Adaptation: Legislators and policy-makers, as well as the legislations and policies themselves, mostly need to catch up with the advancing technological developments, as regulating them frequently requires specialised knowledge. Access to expertise in the field tends to be curtailed by protection extended to the said new technologies in the form of corporate intellectual property. A need has been expressed to legislate upon the area, to confer some control to governmental agencies so that GenAI may be governed and ensured as a public good.
Use of Content Without Consent: Building the GenAI models requires vast amounts of data, which are constantly the result of unlawfully lifting information from the Internet without its owner's permission. This violates Intellectual Property rights and contravenes laws such as the EU GDPR, specifically in aspects like the "Right to b" forgotten", as fed d"ta cannot be removed after training a GPT model.
Unexplainable Models Used to Generate Outputs: The working of Artificial Neural Networks lacks transparency, and it is not possible to work out and arrive logically at the generated output due to the presence of models, their parameters and weights, which are all incapable of inspection yet end up determining the pattern of functioning and serving as the basis for the output. This leads to the inability to remove biases in training data and causes distrust among potential users due to its opaque functioning.
AI-Generated Content Polluting the Internet: The lack of strict regulations and monitoring mechanisms have allowed discriminatory and biased materials generated by AI to spread across the Internet. AI-generated content tends to appear perfect and can mislead learners who rely on it without firsthand knowledge.
Lack of Understanding of the Real World: GenAI tends to merely repeat language patterns found on the Internet and ultimately in its training data, disconnected from the generated subject matter's mematter'seal-world observations, aspects of scientific methods and human and social values. This makes it fail to live up to the trust reposed in it.
Reducing the Diversity of Opinions and Further Marginalising Already Marginalised Voices: The values of the owners and creators of the data get reflected in the output of GenAI, leading it to echo dominant beliefs, statements and ideas that have been repeated a lot. This leads to the voices of marginalised communities being further drowned by not being reflected in outputs due to their insufficient online presence and voice.
Generating Deeper Deepfakes: As image GenAI is trained by the repeated generation and discrimination of images between authentic images and the rendered result and is trained to make increasingly realistic images and reduce the distinguishability between its generated result and original photos, the same capabilities are misused by malicious actors, as it gets increasingly more accessible to create convincing but fake, morphed images, videos and films, to curate their false narrative.
Use of GenAI in Education - Regulation
A human-centred approach to AI calls for the idea that AI is to serve the development capabilities for furthering a future of justness and inclusiveness. As it would need to be guided by human rights principles, human dignity and the promotion of cultural diversity, there is a need for regulation to ensure accountability, transparency and human agency. UNESCO's 202UNESCO'smendat" on the Ethics of Artificial Intelligence proposes a framework of norms for it.
It has been agreed upon that for education, the use of AI technologies is to enhance sustainable development capacities and promote effective human-machine collaboration. AI educational policies adopt an approach that includes the whole government, multiple sectors, and stakeholders. This has been discussed in the 2019 "Beijing Con" census on Artificial Intelligence (AI) and Education".
A need has" been expressed to enable inclusiveness in access to learning programmes, options for personalised and open learning, better access and quality in education, monitoring the learning process, and understanding how to use AI meaningfully and ethically. This has been emphasised in the UNESCO 2022 "AI and Educ"tion: Guidance for Policy Makers".
Steps to"Regulate
Encouraging the development of data protection regulations in countries that do not have them and better implementation of existing frameworks in countries that do.
Development of national strategies, prioritising whole-government action towards the specific issue of Generative AI.
Urgent articulation and implementation of ethics regulations in countries that do not have them, and formation of guiding principles and translation into enforceable laws and regulations in countries that do have them.
Adjustment of existing laws of countries to address the issue of intellectual property issues surrounding the use of data used to train the AI and the status of outputs generated by the AI.
Development of GenAI-specific regulations, such as the Provisional Regulations on Governing the Service of Generative AI (Cyberspace Administration of China, 2023.
Capacity-building, training and coaching for the effective adoption of AI.
Public debates and policy dialogues to foster better understanding and reflection of the long-term implications of GenAI in education.
Key Elements
Government Agencies
National bodies will be established to pave the way for a whole-government approach and sector coordination.
The GenAI framework in place must be aligned with regulations and legislations that especially entail data protection laws and regulations on internet security, among others. Existing rules are to be assessed, and the necessary adaptations will be made as and when needed.
Cooperation between sectors is necessary for ensuring better trustworthiness among GenAI models. Open-source ecosystems can be utilised to share computing resources and pre-training datasets. The practical application of GenAI across industries must be fostered to ensure public benefit.
Principles and processes are to be set in place for assessing and categorising the safety, security and efficacy of Gen AI services, with the classification of Gen AI applications based on risk levels and strict regulations for those classified under high risk.
Laws should be introduced to protect personal information while using GenAI and against unlawful and unauthorised data storing, sharing and profiling.
GenAI is to be made age-restricted, with age verification measures and a set of accountabilities defined for the providers of GenAI and the guardians of child users.
The protection of national data ownership is to be brought forth through legislation and regulations on GenAI providers to promote mutual benefit cooperation.
Providers of GenAI tools
Human accountabilities must be ensured to promote core values and implement lawful and ethical practices and purposes.
Data and foundation models with proven and demonstrated trustworthiness and legal sourcing with adherence to intellectual property laws are to be mandated.
Boundaries must be established to protect against discriminatory or biased content and offensive and false outputs.
Transparency must be practised concerning the various data sources used for the GenAI models, with explanations submitted to public governance agencies.
Gen AI content and output are to be labelled as machine-generated.
Throughout the life cycle of the GenAI, an assurance is needed as to its security, robustness and sustainable service.
To ensure rational and responsible decisions, contextual specifications will be made regarding the scenarios, purposes, and target audience.
Limitations of the GenAI are to be specified, and guidance is to be accorded to prevent predictable harm, including addiction and over-reliance on generated content.
A mechanism for addressing and remedying the users' composers through timely action is to be introduced.
Illegal and unethical use, such as for disinformation and hate speech, is to be monitored and reported in furtherance of cooperation with public governance agencies.
Institutional Users
Mechanisms are to be established to monitor the data, algorithms and outputs within the institution, and regular audits and assessments are to be conducted to ensure ethical standards and user data protection, along with eliminating inappropriate content through automatic filtering.
The GenAI systems and applications are to be classified under national classification mechanisms as well as institutional policies, and it is to be ensured that the GenAI aligns with the locally validated ethical frameworks to prevent predictable harm to the institution's institutions.
The potential effects of using AI in the long term are to be reviewed, and the potential implications it might have over the development of critical thinking and creativity are to be considered.
Implementing age restrictions concerning the independent use of AI within the institution is to be considered.
Individual Users
Awareness of the terms of reference, the obligations stipulated, and the applicable laws and regulations behind the agreement are necessary.
Responsible, ethical use of AI is to be carried out, ensuring no harm or violation of the reputation or rights of any person or persons.
GenAI applications that violate regulations must be notified to governmental regulatory agencies.
Policy Framework
Promotion of Inclusion, Equity and Diversity
Those individuals who lack internet connection and data access are to be identified, and digital competencies are to be undertaken to bring down the barriers to ensure inclusiveness, equitable access and universal connectivity. Sustainable funding mechanisms for AI-based tools will be established for learners with disabilities and special needs. The use of GenAI is to be advocated as support for diverse groups of learners across locations, age groups and backgrounds.
Criteria are to be developed for the validation of GenAI systems for the elimination of gender bias, marginalised groups being discriminated against and hate speech embedded in data and algorithms.
Inclusive specifications and requirements will be developed to include data in multiple languages, considering various linguistic and cultural diversities. Bias against indigenous languages and their speakers, including the removal of such languages, are to be prevented strictly. The promotion of dominating languages and cultural norms by providers is to be avoided.
Protect Human Agency
The learners are to be informed about the types of data collected by AI, its use, and its impact on their education and broader life aspects.
The learners' motivation for growth and learning is to be preserved, and even in the context of GenAI systems, human autonomy over research teaching and learning is to be reinforced.
How the GenAI is put to use should be that there is no deprivation of opportunities to develop cognitive abilities and social skills through real-time firsthand observation, experimentation, discussion and independent logical reasoning.
Learners are to be prevented from getting overly addicted or reliant on GenAI and be ensured sufficient social interaction and exposure to human-produced creative output.
The use should focus on minimising academic pressure rather than the opposite of the use of GenAI tools.
The views of researchers, teachers and learners are to be gathered, and the feedback is to be used to make decisions on deploying specific GenAI tools on an institutional scale. The impact of AI methodologies is to be critically evaluated.
Delegating human accountability, particularly in high-stakes situations, must be avoided.
Monitor and Validate GenAI Systems for Education
Validation mechanisms are to be established to assess GenAI systems for biases, and verification is to be made regarding the AI's training of diverse and representative data.
The issue surrounding informed consent is to be addressed, mainly where the learners fall under a vulnerable category, such as children.
Regular Audits are to be conducted to ensure against deepfake images, fake news and hate speech in outputs, and swift action is to be taken to mitigate and eliminate any such identified instances of inappropriate content.
Before the adoption of AI by educational institutions, strict ethical validation is to be carried out, following an ethics-by-design approach.
It is to be ensured that GenAI applications do not predictably harm students, are effective educationally, and are suitable for the ages and abilities of the learners and the pedagogical principles based on relevant knowledge domain before the application adoption.
Develop AI competencies, including GenAI-related skills for Learners.
Government-sanctioned AI curricula for school education and technical and vocational training are to be committed to. It is to be ensured that the curricula cover the impact of AI on lives, ethical considerations, understanding of algorithms in an age-appropriate manner and skills for the creative use of AI tools, including the said GenAI applications.
Support is to be provided to higher education and research institutions for developing AI talent through enhancement programs.
Gender equality is to be promoted in the advanced AI competency development to create a gender-balanced pool of IT professionals.
Forecasts of various job - shifts at a national and global level among different sectors are to be developed. Based on the prospective demand, skills at all levels of education are to be enhanced.
Special programs are to be provided for older workers and citizens to learn new skills and adapt to new environments in the evolving nature of work due to AI.
Build Capacities for Teachers and Researchers to Make Proper Use of GenAI
Local guidance is to be formulated based on testing for the assistance of teachers and researchers in the navigation of the GenAI tools that are widely available, and the design of new domain-specific AI applications is to be directed.
The teachers' and teachers' values must be upheld while utilising GenAI. Their roles in facilitating high-order thinking, human interaction and fostering human values are to be recognised.
The value orientation, knowledge and skills teachers would be required to utilise the GenAI systems effectively are to be defined. The creation of GenAI-based tools to support learning and professional development is to be enabled.
The competencies the teachers require to understand and effectively use AI for teaching goals, learning and professional development are to be continually reviewed. The emerging AI-related values, skills and understanding will be integrated into competency frameworks and training programs for in-service and pre-service teachers.
Promote Plural Opinions and Plural Expressions of Ideas
It is acknowledged that while GenAI may be quick, it is often an unreliable source of information. It is to be understood that plugins and Learning Language Models (LLM)-based tools effectiveness lacks robust evidence.
The critical evaluation of the responses generated by GenAI by the learners and researchers is to be encouraged. The potential undermining of diverse perspectives and minority opinions due to GenAI repeating established opinions are to be recognised.
Opportunities for empirical learning through trial and error, experimentation and observation of the natural world are to be provided sufficiently to the learners. A holistic and practical understanding is to be targeted through hands-on experiences and GenAI-generated information.
Test Locally Relevant Application Models and Build a Cumulative Evidence-base
The design and adoption of the GenAI are to be strategically planned to avoid passive and non-critical procurement processes.
GenAI users are to be encouraged and incentivised to target learning options that are exploratory, open-ended and diverse to foster their innovation and creativity.
Evidence-based use cases of AI in research and education are to be tested and scaled up, with educational priorities being the focus.
GenAI is to be guided towards innovation in research, leveraging computer capability, and improving research methodologies by large-scale data and GenAI outputs.
The social and ethical implications of incorporating GenAI into research will be reviewed to ensure ethical use.
Specific Criteria based on evidenced pedagogical research and methodologies are to be established. An evidence base is to be built for GenAI to support inclusive learning opportunities effectively, meet learning and research objectives, and promote linguistic and cultural diversities.
Iterative steps are to be taken to strengthen evidence about the social and ethical impact of GenAI and to ensure continuous responsibility and improvement.
The environmental costs of leveraging AI technologies at scale, including the energy and resources required to train GPT models, will be analysed. Sustainable targets are to be developed for AI providers to mitigate environmental impact and minimise climate change.
Review Long-Term Implications in an Intersectoral and Interdisciplinary Manner
AI providers, educators, researchers, and representatives of parents and students are to collaborate to plan system-wide adjustments in GenAI-related curriculum frameworks and methodologies of assessment.
Educators, researchers, learning scientists, AI engineers and other stakeholders must be brought together for intersectoral and interdisciplinary expertise. The long-term implications of GenAI on aspects such as learning and knowledge production, research and copyright, curriculum and assessment, human collaboration and social dynamics are to be examined.
Timely advice will be provided to inform iterative updates of regulations and GenAI-related policies in education and research.
Conclusion
A human-centric approach is needed concerning Generative Artificial Intelligence tools. This is especially necessary to be a trustworthy tool for researchers, learners and teachers. There is a need for a rigorous review of the potential of GenAI to transform established systems and foundations in education and research. Careful consideration is crucial to harness AI's potential and AI's educational technologies, ensuring the enhancement of human capabilities and contribution towards a sustainable and inclusive digital future.
Comments
Post a Comment