What the Index Has to See

On May 29, the dean of Boston University's Wheelock College and the executive director of Just Horizons Alliance, a small Boston applied-research nonprofit, signed a three-year partnership at Boston Tech Week to build something the field has been quietly missing.¹ They are calling it the AI Ethics Index for K-12 Education. Axios Boston wrote it up the following Monday in a story that landed without much noise, the way the most important infrastructure stories usually do.¹

The Index will measure nine dimensions. Model design and development. Fairness. Privacy and data stewardship. Transparency. Knowledge and attribution. Human-AI interaction. Safety and security. Societal impact. Governance and accountability.² The output of an assessment will not be a single number. It will be what the team calls a "structured integrity profile," a drill-down visualization that lets a superintendent see where a tool is strong, where it is weak, and where a vendor is making a claim the evidence does not yet support.²

Eight of those nine dimensions have settled literatures behind them. Privacy law has been adjudicated since FERPA. Fairness audits have a decade of methodology to draw on. Transparency, attribution, governance. Each has a journal, a working group, a body of practice.

The ninth is the one that will decide whether the project amounts to anything. Janet Kang, the alliance's executive director, named it in her launch quote. The Index, she said, is meant to evaluate "whether they strengthen or undermine the conditions students need to learn, develop, and thrive."³

The Quiet Problem of Measuring Learning

The trouble with the ninth dimension is that the field has not, for most of its existence, had a steady way to see learning happen. It has had a way to see outcomes. Test scores. Grade-point averages. Graduation rates. Each of those is the dust the practice leaves behind. None of them is the practice.

An AI tool can lift a test score by handing a student the answer. It can lift a grade by writing the essay. It can lift a graduation rate without ever strengthening the conditions Kang described. Outcomes are the easiest thing in education to game, which is why a serious index of learning alignment cannot rest on them.

What it would have to rest on is the record of practice. The draft a student kept and the one she abandoned. The question she asked the tutor that turned out to be the wrong question, and the better one she asked an hour later. The pause she took before she answered. The minute thirty-seven breakthrough she did not know was coming when she sat down at minute thirteen. The paragraph she would have written without the tool, and the paragraph she wrote with it. The gap between those two is the only place where the question of strengthen or undermine can be answered with evidence.

A Future of Honest Evaluation

If a K-12 evaluation index is going to do what its authors are asking of it, school systems will need to develop the practice of keeping that record. Not for surveillance. Not for grading. For inspection of the kind Kang and the Wheelock team are describing. To put a tool on the table, ask whether it strengthened a child's practice or quietly replaced it, and answer with something other than the finished homework.

That will change what the phrase "good ed tech" means. Today it often describes a tool that produces clean outputs in a tidy interface. Under a serious Index, it would describe a tool that leaves a richer trail of the student's thinking than the student would have left alone. Two tools could produce identical final essays and earn very different scores on the ninth dimension, because one would have surfaced the practice and the other would have absorbed it.

This is also the part of the Index that will be hardest to staff. Fairness, privacy, governance have ready-trained reviewers. Learning alignment requires people who can look at a trail of practice and recognize the shape of a student's thinking. That is closer to what good teachers already do at a Wednesday morning conference than to anything an auditor learns in a certification course. The Index will, in effect, ask the field to put teacher judgment back at the center of the evaluation of tools meant for teaching.

What an Honest Index Has To See

This is the part Koan has been quietly built for. Not as another tool to be scored against, but as a way of giving classrooms the record an honest Index would need to read. The revisions. The pauses. The exchanges a student has with Aidan that surprised her halfway through. The shape of how a sentence got written, not only the sentence that ended up on the page. Outcomes are still there at the end. So is everything in front of them.

The AI Ethics Index is the kind of project the field has needed for two years. It says, gently and with the right amount of humility, that the era of believing a vendor's slide deck about learning impact is ending. The work ahead is harder. It involves looking. It involves choosing to look at the right thing. And it requires, if the ninth dimension is going to mean anything at all, classroom tools that make the right thing actually possible to see.

If your district had to defend the AI it bought a year from now, what record will you wish you had been keeping?