Towards a Framework for Sustainable and Ethical Use of Artificial Intelligence in K-12 Education

Defining rights and protections for the most critical stakeholders in AI-based education: students, parents and teachers

Michael Copass, Science Educator

July 4, 2024

San Diego

This framework is a living document, meant to be the opening of a discussion, not the end. I hope it provokes a discussion of what could be, and should be for AI uses in education. Never hesitate to dream of better.

AI LITERACY GOALS FOR ALL: PARENTS, STUDENTS, AND TEACHERS

Given the way the AI technical world is undergoing rapid evolution, both minor-aged students and their parents and guardians need to be provided with the education and skills necessary to understand how AI works, and how to integrate AI usage into the child’s ethical learning process, preparing them for more AI in college/higher education and the “workplace of the future.” Solid AI literacy is also foundational for AI-involved homeschooling, a growing route. As the main engine of education, schools should take the lead in teaching AI literacy to all stakeholders: teachers, students, and parents. Well-developed UNESCO AI literacy frameworks may be used. More evidence is needed to support the role of AI literacy in modern learning.

AI SHALL BE EMPLOYED AS A HUMBLE CLASSROOM SERVANT TO ADVANCE, NEVER DEGRADE, STUDENTS’ CRITICAL THINKING SKILLS

With AI supporting content and knowledge acquisition, teachers can increasingly emphasize the higher level critical thinking skills needed to train skilled and reflective learners capable of higher inquiry skills, such as are referred to in #BloomsTaxonomy. Moreover, with AI as a personal tutor building knowledge base, students can develop claim-evidence-reasoning scientific reasoning, on their way to becoming informed citizens capable of parsing misinformation. If done right, AI may even help instill a delight in lifelong learning among young scholars. Given the massive importance of these cognitive skills, AI shall always be used to develop and strengthen, never degrade, critical thinking and higher level cognitive functions in children. IF THINKING SKILLS ARE BEING ERODED, FULL STOP: and reread this document.

AI SHALL BE EMPLOYED AS A HUMBLE CLASSROOM SERVANT TO ENHANCE, NEVER DEGRADE, STUDENT CREATIVITY

As a skill particular to humans, student creativity in all forms, art, language arts, math, philosophy, science, etc., can be now emphasized to a greater extent - with AI tools doing the support work of knowledge acquisition and personalized tutoring. AI shall be used only to enhance, and inspire, and never degrade, a child’s creativity. More evidence is needed to explore how AI can support human creativity. (IF YOU FIND AI IS ERODING CREATIVITY, FULL STOP: and reread.)

SAFETY STANDARDS FOR AI

The United States Food and Drug Administration (FDA) agency requires rigorous, defined safety data for any new candidate medicine proposed by a pharmaceutical corporation. It makes common sense that AI, like a candidate medicine or a new model of child safety seat, should pass defined safety standards. AI tools should be re-evaluated frequently for safety, as rapid AI evolution is predicted. Absolutely no harvesting of a child’s data by Artificial Intelligence shall ever be permitted, under stringent penalties. In the interests of psychological health and safety of children, no emotional deception of children by AI is allowed (see below.) The transparency of the dataset the AI is trained on, and its algorithms, shall be open to all who inquire, particularly parents. Abundant studies and evidence is needed to establish AI safety in numerous areas. 

EFFICACY STANDARDS FOR AI

AI should also be subject to meeting standards of effectiveness in advancing student learning. Just as the FDA imposes safety standards to protect the public, the same agency also insists on rigorous clinical trials to prove that the new medicine is actually effective. The candidate drug must show that it lives up to its marketing claims. backed with data showing significance. We apply the same standard here to Educational AI tools, and there is an opportunity for numerous studies to provide evidence of effectiveness.  If AI tools are currently marketed with the message or implication that they can advance student learning, or make similar claims, here is a simple standard for all AI/EdTech vendors to follow:

Prove it.

AI & MACHINES ARE NOT ALLOWED TO PROMOTE EMOTIONAL DECEPTION

AI shall always identify itself to children as a non-human machine. For example, the chatbot will greet the student with “Hi, I’m a Machine. How can I help you?” No AI used on children shall be permitted to approximate a human voice, face, likeness, or personality. Avoid any risk, however small, of emotional manipulation of a child by AI. Machines are servants to education, not “friends” and Artificial Intelligence constructs are never permitted to groom children to be electronic pals or confidantes. Children’s brains have less completely developed frontal cortices as compared to adults, and children may be more easily drawn into emotional relationships with a realistically human-appearing machine. No frauds. No fakes. Be honest.

AI TRAINING DATASET: MAXIMUM TRANSPARENCY

AI tools used in the classroom must provide complete and publicly accessible datasets on which the AI was originally trained. All subsequent training updates will be published. Parents have a right to know what material their child’s AI is quietly learning and studying. If for-profit AI training datasets are used (private sector AI) they should be scoured for potential private personal data hidden within, which many experts suspect is secretly harbored. 

Yes, Solutions are challenging. An end-around these difficult transparency issues is to create a publicly funded LLM. Thus instead of a “black box” which creates huge headaches. a publicly funded LLM could be created for the entire United States educational establishment to use. I could be built like the Next Generation Science Standards (NGSS) were created, via an intense and fruitful collaboration amongst educators. Up to this point, no evidence exists that shows for-profit, corporate AI will make any training data. This cannot work for education.

COSTS, COMPETITIVE PRODUCTS AND CLIMATE CHANGE

Purchase costs of all AI and EdTech to School Districts ,on behalf of children, shall be published and accessible, stimulating healthy competition and reducing costs to the educational system’s budgets. Given the massive amount of often carbon-producing energy needed to train and operate AI, full carbon accounting of all school-related AI processes and uses will be published. This is essential to promote green or carbon-conscious ESG competition between tech firms, allow for “greener” choices in AI providers, and to mitigate AI-driven carbon dioxide emissions that impact global climate change.

MACHINES SHALL NOT REPLACE HUMAN TEACHERS IN THE CLASSROOM

To provide the interactions, empathy, and critical emotional support that only humans can provide to developing children, the in-class roles of teachers, teacher’s aides, and special education personnel will be carefully preserved. Yes, artificial Intelligence will be utilized in support of advancing a teacher’s effectiveness; a “force multiplier” of a teacher’s skills. But AI can not replace the teacher, as it cannot do much of what they provide. We say forcefully that AI shall never be developed with a view to supplanting or replacing the varied roles of human teachers in the classroom. Importantly, AI and EdTech budgets shall never be commingled with the teacher payroll; they are not in the same categories. Increasing tech expenditures shall never directly lead to decrease in teacher staffing.

INFORMED CONSENT AND OPT OUT

As K-12 students are overwhelmingly minors ages 5-17, children and their parents may opt in or out of what the EdTech community has branded “transformative technology.” “Transformative” surely implies parental involvement, and informed parental consent shall serve as a guide and foundation, frequently updated. No child shall be made to use AI which has not been shown to be both safe and effective. The family is the key stakeholder in the “AI + School + Learner” equation, and will be treated with utmost respect as the critical client.

In closing: already, over 2000 AI tools have already flooded the education technology (EdTech) space. Thus, a carefully crafted pro-student, pro-family, pro-teacher, and pro-education framework is essential as a guide, a set of guardrails, and an ethical touchstone. It should be the product of active debate,  and happen at the opening of the AI era. We have but #OneChance to get AI right in schools, and this document is intended to support the construction of thoughtful and robust guardrails to protect children, parents and teachers. Continuous discussion of these proposals, and iterative improvements will support the safe and effective integration of AI into the classroom according to a sound ethical framework.  

AI will likely evolve faster than we can imagine today; we must prepare all stakeholders to learn, pivot, and grow, quickly.  The goal is that education may experience the maximum benefits of this powerful technology, that teachers may instruct, model and direct children in its ethical use inside and outside the classroom, all while protecting children and families from potential harms. The pace of change may be exhausting; therefore a pre-existing framework, and vigilance, is essential in safeguarding children’s health and success as learners, for parents, teachers, and the educational process itself in a rapidly evolving AI-saturated world.