How will AI change education? Will the change be revolutionary? Instead of speculating, I will base the answer on a concrete case currently in commercial use. This is an autonomous AI system that screens and corrects a neurodevelopmental disorder, dyslexia.
Dyslexia is a reading and learning disability that is addressed primarily by the public school system. Dyslexia affects 1 in 5 children. The way it is handled in schools presently illustrates well why the education industry is ripe for revision.
Assessment
Assessment is a resource-heavy endeavor in the public school system. Of course, all stakeholders–teachers, parents, administrators, taxpayers–need to know how students are faring from public investment in their education. The problem is how assessment is done presently.
Integrated Assessment-Instruction
At several points in the school year, students and teachers have to drop everything for testing. At the very least, there are annual state tests for reading and math, quarterly standardized tests for the same, and school tests on all subjects at every term.
But why do students and teachers have to drop everything to accommodate tests? It is only so because assessment is presently separate from instruction. What if we can integrate assessment and instruction?
In the dyslexia example mentioned above, every session with the AI system is both evaluative and developmental. In interacting daily with a student through a game interface, the AI system collects and analyzes information about their brain processing and generates new game activities to correct problem areas. The reading difficulty associated with dyslexia is due to problems processing language in the brain.
Thus, every game session serves both evaluative and developmental purposes. Additionally, with gamified assessments, the student does not even feel that they are taking a test.
Why do schools test students so often? Because today’s tests are merely static snapshots of a pupil’s performance on a few sample items at a single timepoint.
In contrast, if every daily interaction of a student with an AI system is evaluative, we can obtain a dynamic, comprehensive picture of their abilities, weaknesses and changes. With this evolving blueprint, instruction or intervention can be considerably more sensitive to this particular child and consequently more effective.
No Datum Left Behind
What do schools do with all their students’ test reports? Firstly, schools do not always get assessment outcomes immediately to guide instruction. State test results may be released several months or even a year later.
Secondly, the traditional approach of having teachers interpret assessment data for each student and modify or differentiate instruction accordingly is logistically and economically unfeasible.
Do the math:
1 annual state test + 4 quarterly tests = 5 tests
5 tests x 2 subjects (reading and math) = 10 test reports
10 test reports x 20 pupils per class = 200 test reports
This figure for one teacher for a single school year does not include other assessments such as term tests for report cards, quizzes etc.
A student with a disability such as dyslexia would have even more evaluation data that require interpretation and application. Students in special ed have Individualized Education Plans (IEPs), which are crafted one by one by a team of specialists and teachers. These IEPs contain the results of even more evaluations, narratives and other data.
The U.S. is experiencing a shortage of teachers, acutely in special ed. With traditional means, the more stakeholders demand accountability and documented outcomes, the more impossible the job becomes.
While 20% of the population have dyslexia, only around 15% of the student population in grades K-12 are in special ed for all disabilities. This means that, at best, barely half of students with dyslexia get special services. Resources at school are in short supply.
It is not humanly feasible to do what is required to meet all students’ needs: appropriate and timely evaluation, thorough interpretation of results and differentiation of intervention for everyone. These are tasks that can be computerized and automated, as done with AI-driven dyslexia intervention.
In short, AI can scale these laborious tasks to serve every child in need on demand. Realtime assessment provides the instantaneous feedback needed for realtime response.
No child is left behind when no datum is.
From Many, One
From One, Millions
For anyone unfamiliar with educational assessments, the plethora of test instruments for K-12 literacy alone may seem like an alphabet soup: WJIV-R, WIAT, PIAT-R/NU, WRMT, GORT-4, NWEA MAP. Some test reports show scaled scores; some standard, composite or stanine scores.
Parents have asked me, Why are there 27 reading or lexile levels when there are only 12 grades? And what are lexiles anyway? Some of these parents are teachers themselves, just not in Reading or English Language Arts.
Consider that school superintendents and teachers move from district to district in their careers. Just for identifying students at risk of reading difficulty, at least a dozen screeners are available in the market. While educators take the time to get oriented to the assessment tools of their new workplace, important decision-making may have to wait.
Imagine this old school system this way: All the students in the district are in a stadium. To assess them, administrators have to decide which tunnel out of many to direct each student to. Bottlenecks occur in the traffic flow and it takes time to usher all students through the process.
Instead of this messy process, AI can handle all students all at once. From the educator’s perspective, all students go through a single portal, the user interface for the AI program. But actually, behind this simple user interface, AI is generating millions of person-specific paths for students, based on each pupil’s learning profile.
The technical complexity is not something school personnel need to understand or manage at all. In the AI universe, users simply use. And AI products should be as easy to use as possible.
The Individualization of Education
Schools and ed tech companies have been talking about personalization for decades now, but AI can customize instruction to a much more granular level than just adapting learning to preset stages. In the new AI world of education, there are no stages. Some subject matter such as math may be suited to step-wise acquisition. But skills and abilities need not be.
Think of 3 tiers. The middle tier is skills acquisition, like learning spelling rules or reading. Skills support the top tier, the learning of content like science and history.
The bottom, foundational tier is brain processing. Brain processes support skills acquisition.
Teachers cover the middle and top tiers. AI opens up a way to reach the foundational tier, brain processing, for the first time. The AI system for dyslexia improves language processing in the brain so that the student can acquire spelling, word decoding and other skills necessary to read well.
This AI impact can be far-reaching. Teacher instruction becomes effective for formerly failing learners. Longterm special services are no longer necessary.
Resources can then be re-allocated, say to enrichment programs for everyone. School budgets, and hence taxes, can be less costly. U.S. schools presently spend over $120 billion a year on special ed, the largest category being dyslexia.
The Humanization of Education
AI humanizes education by enabling greater interaction outside of the technology. AI does so by freeing up a precious resource: time. Consider what this means from the viewpoint of the teacher and parent.
When AI assumes burdensome tasks that computers do well, teachers are liberated to carry out what they do best. With traditional assessments, they had to get trained and sometimes certified to administer the testing, conduct the tests, score and interpret the results.
With AI-run assessments, teachers can then spend more time and energy on helping each child meet their full potential. Many teachers likely chose their profession to help shape a young person’s life. While some students are engaged with an AI program, teachers can get to know the rest better, their talents and aspirations, and motivate them.
Teachers will also have more time for concerned parents, to communicate with them about their children’s problems and progress. Or plan ahead jointly.
Indeed, parents will also have a new source of information to understand more fully what is happening to their children at school. Parents can ask the AI system questions in plain language. Based on their children’s assessment data, the system will generate the answer, also in plain language.
Parents can ask the system as many times as they want, without fear of being annoying. They may ask what they always wanted to know but were too embarrassed to ask. Like what lexiles are.
Prediction
To predict the future of AI in education, look for endemic problems in the school system. Sooner or later, someone is going to come along with breakthrough technology and solve them.
