How Do You Assess Inquiry-Based Learning? Designing Rubrics and "Process Assessment" in the AI Era

This is Hamamoto from TIMEWELL.

How should inquiry-based learning be assessed? In Japanese high schools, the Period for Inquiry-Based Cross-Disciplinary Study is a required subject worth three to six standard credits¹, yet it cannot be scored with a paper test, and the official student record calls not for a grade but for teachers to "describe concisely in writing what abilities the student has developed"². What assessment can lean on are the three perspectives set out under the national Courses of Study — knowledge and skills; thinking, judgment, and expression; and an attitude of proactive engagement with learning — together with the criteria and rubrics each school builds for itself. On top of this comes generative AI, which is quietly breaking any assessment that looks only at the finished product. MEXT's guideline explicitly lists submitting AI-generated output almost unchanged as one's own work among its inappropriate examples³. In this article, I walk through why inquiry assessment is structurally hard, how the three perspectives map onto the inquiry process, how to build a rubric with concrete descriptors, how to assess the process through AI chat logs and trial-and-error records, and how to connect it all to presentations and portfolios — grounded in primary sources from MEXT and the National Institute for Educational Policy Research (NIER).

Our core business at TIMEWELL is hands-on AI support for companies, but when I talk with teachers through our work with schools, the first worry they raise about inquiry is theme setting — and the one that surfaces once we know each other a little better is assessment. Three key points up front.

• Inquiry assessment is hard not because teachers lack skill, but because the system itself makes each school define its goals, content, and perspectives. That is exactly why schools that build their own yardstick — a rubric — are the ones that find relief.
• The foundation is the three perspectives in the Courses of Study. In particular, "attitude of proactive engagement with learning" is defined as having two sides — persistent effort and attempts to self-regulate one's learning — and is not something to grade on impressions of enthusiasm.
• In the generative AI era, records of the process take the leading role. Building AI chat logs and trial-and-error records into assessment deters outsourcing to AI and improves assessment validity at the same time.

Why is inquiry so hard to assess? Knowing the structure settles the mind

Let me make one thing clear at the outset. If you agonize over inquiry assessment, it is not because you lack experience. The system is built in a way that guarantees the agonizing.

The first reason is that there is no right answer. A math exam can be marked right or wrong, but inquiry themes differ from student to student. You cannot compare a student who studied local disaster preparedness with one who worked on revitalizing a shopping street using a single yardstick of "quality of output." Inquiry also frequently ends without a neat conclusion, and if you grade on how impressive the conclusion looks, you mass-produce students who choose safe themes that wrap up nicely.

The second reason is that the yardstick itself is something each school must build. For the Period for Inquiry-Based Cross-Disciplinary Study, the Courses of Study present only a broad goal; specific goals and content are to be defined by each school. On assessment perspectives too, the MEXT notification that defines the student record format says schools record "the assessment perspectives each school has defined for itself"² — the state does not hand out a detailed scoring table. No textbook, no nationwide standard. Compared with other subjects, this is a course where you cannot mark anything until you have written your own marking scheme.

The third reason is the absence of a numerical grade. The student record takes a written description alongside the learning activities and perspectives, not a five-point grade². The absence of numbers looks like a mercy, but it is a trap. Without the forcing function of a score, an undesigned assessment drifts toward "they worked hard, so a good evaluation" — an impression you cannot defend when a student or parent asks for the grounds.

Given this structure, I believe there is only one sensible resolution: build your school's own yardstick and show it to the students. The concrete tool for that is the rubric, and its foundation is the three perspectives.

The foundation: the three perspectives, mapped onto the inquiry process

A March 2019 MEXT notification organized learning assessment across all subjects into three perspectives: knowledge and skills; thinking, judgment, and expression; and an attitude of proactive engagement with learning². Here is what the state says each perspective means for the inquiry period, side by side with where in the inquiry process that ability becomes visible²⁴.

Perspective	Officially stated purpose (summary of the original)	Where it shows in inquiry
Knowledge and skills	Acquiring the knowledge and skills needed to find and solve problems through the inquiry process, forming concepts related to the task, and understanding the meaning and value of inquiry	Scenes where students connect findings to prior knowledge; command of methods such as interviews or basic statistics
Thinking, judgment, and expression	Finding a question out of one's own relationship with real society and real life, setting one's own task, gathering information, organizing and analyzing it, and summarizing and expressing it	Reworking the question, choosing sources, the line of analysis, the structure of reports and presentations
Attitude of proactive engagement	Engaging with inquiry proactively and collaboratively, drawing on one another's strengths, creating new value, and seeking to realize a better society	Recovering from dead ends, revising plans, collaborating with peers, the substance of written reflections

Notice that the purpose of "thinking, judgment, and expression" is the inquiry process itself. The Courses of Study write into the goal of the inquiry period the sequence of "finding a question from one's relationship with real society and real life, setting one's own task, gathering information, organizing and analyzing, and summarizing and expressing"⁵. Task setting, information gathering, organizing and analysis, and summarizing and expression — each of the four scenes of inquiry is, as it stands, an assessment scene. You do not need to prepare a special test; the worksheets and records each scene produces are your assessment materials.

The other perspective that is widely misunderstood in practice is the "attitude of proactive engagement with learning." It is not graded on hand-raising counts or visible enthusiasm. The Central Council for Education's report defines this perspective as having two sides to assess: "the side of attempting persistent effort toward acquiring knowledge and skills or developing thinking, judgment, and expression," and "the side of attempting to regulate one's own learning in the course of that persistent effort"⁶. Persistence, and self-regulation. Neither appears in a single class period's demeanor; both only ever surface in records of trial and error accumulated over time. This is precisely what connects to the "process assessment" I discuss below.

As an aside, Yoichi Ito, dean of the Faculty of Entrepreneurship at Musashino University, told an audience of high school career-guidance teachers that "the essence of entrepreneurship education is to keep asking, 'What do you want to do?'"⁷ I understand attitude assessment in inquiry the same way: at bottom, it is the work of reading from the records how a student's own will moved. The movement of will leaves no trace in a finished product. That is why the records need designing.

How to build a rubric: five steps and sample descriptors

A rubric is a table that crosses assessment criteria (what to look at) with achievement levels (how far the student has come), with each cell describing in words what a student at that level looks like. Five steps.

1. Derive criteria from the stated purpose of each perspective. NIER's reference materials show a method of converting the wording of the Courses of Study into criteria by changing the sentence endings into "is doing"-style statements⁴. You do not need to invent from zero.
2. Narrow down which scenes of the inquiry process you will observe. Trying to see every criterion in every lesson collapses. Two or three criteria per unit is realistic.
3. Write the descriptors level by level. First write the "satisfactory" B profile concretely, then write A as what is added to B, and C as what is missing from B.
4. Hand the rubric to students at the start. NIER's reference materials also note that schools are expected to create opportunities to share the assessment policy with students in advance, both to raise the validity and reliability of assessment and to give students a clear view of the learning ahead⁴.
5. Revise through use. Have several teachers assess the same worksheet, and rewrite the descriptors wherever judgments split. A rubric is not built once; it is grown.

Words alone are hard to grasp, so here is a sample set of descriptors for "task setting," the first gate of inquiry, under the thinking-judgment-expression perspective.

Level	Task-setting profile (sample descriptor)	Evidence
A (fully satisfactory)	Poses a question from both its connection to real society and its relationship to oneself, narrows it to a size that can be verified through investigation, and explains a hypothesis and a verification method in one's own words	Theme-setting sheet, plan presentation
B (satisfactory)	Poses a question from a theme of interest, ties it to either real society or oneself, and can sketch the steps of the investigation	Theme-setting sheet
C (needs support)	Has not distinguished the theme (the subject) from the question (what is to be clarified), and cannot explain in one's own words what they want to investigate	Theme-setting sheet, dialogue notes

There is only one real knack to writing descriptors: use words that bring an action to mind. Phrases about quantity or impression, like "works enthusiastically" or "researches thoroughly," split judgments between teachers and tell students nothing about what to do. "Explains a hypothesis and a verification method in one's own words" lets student and teacher look at the same goal. Incidentally, if you think through the support for C-level students in advance, the rubric becomes a teaching tool rather than a sorting tool. This is exactly what NIER means by its banner phrase, "the integration of instruction and assessment"⁴. If many students stall at theme setting, having angles ready helps; I have collected AI-centered ones in the 50 AI inquiry themes article.

In the AI era, assess the process: deterring outsourcing and improving assessment in one stroke

Here is the heart of the matter. The spread of generative AI has quietly broken a premise of inquiry assessment: overnight, anyone can produce a plausible report.

The "Guideline on the Use of Generative AI in Primary and Secondary Education (Ver. 2.0)," published by MEXT on December 26, 2024, lists among its inappropriate examples "submitting output generated by AI almost unchanged as one's own work for contests, reports, or essays"³. The direction is clear. But as a practical matter, prohibition alone does not hold the line. Telling AI-written text apart from the finished product gets harder every year.

I believe the axis of the response should shift from detection to design. Concretely: widen the object of assessment from the product to the records of the process, and declare that up front. In fact, the same guideline spells out the hints in surprising detail. Where students use AI to fill gaps in a draft during project research, it suggests "having students attach the record of their exchanges with the generative AI as reference material, and having them clearly cite sources and references," along with "having students note the name of the generative AI tool, the prompts entered and outputs, and the dates," and, where submissions feed into assessment, "creating opportunities for oral presentation to the whole class or in groups"³. Have students hand in the process, and have them speak to it. Read plainly, the national guideline is recommending process assessment in all but name.

So what does a record of the process look like? These are the three I have found most workable in our collaborations with schools.

Process record	What it captures	Perspective it serves
AI chat log (history of prompts and outputs)	How the question evolved, how concrete the instructions were, whether outputs were doubted and checked	Thinking-judgment-expression; knowledge and skills
Trial-and-error record (weekly reflection sheet)	Dead ends and recoveries, plan revisions, persistence	Attitude of proactive engagement
Source list and fact-check record	Whether primary sources were consulted, whether AI output was verified	Knowledge and skills

The elegance of this design is that it works twice over. First, outsourcing becomes structurally a losing move. A student who dumps the task on AI has a chat log that reads "write my report" and ends there, and a reflection sheet with nothing in it. Once the process is declared to be the object of assessment, outsourcing is the same as forfeiting your assessment materials. Conversely, the log of a student who argued with the AI over many rounds, caught an error in its output, and rebuilt the question shines as a fossil record of thinking. And as we saw above, the attitude perspective is defined as the two sides of persistence and self-regulation⁶ — neither of which is ever visible in a finished product. Process assessment, in other words, is not a compromise forced by AI; it is a return to the direction the national assessment framework has pointed at all along.

You may not have time to read every log. You do not need to. Have students write on the reflection sheet "the best question I put to the AI this week, and how I verified the output," with the log attached as supporting evidence. Read the sheets; spot-check the logs. That level of operation is enough. For the classroom design itself of having students work with AI, see the practical article on high school inquiry-based learning and generative AI.

Close the loop with presentations and portfolios

Records of the process do not exert their force as loose scraps of paper at term's end. The closing move is to connect them to a portfolio and a presentation event.

NIER's reference materials include the case of a high school where students distill three years of inquiry into a graduation thesis, present it to first- and second-year students and community participants, and then reconstruct their own thinking in light of the feedback. In that case, everything from the theme-setting sheet to the end-of-unit reflection portfolio, and even remarks made in career counseling, serves as assessment material⁴. Bind the products and the process records into one volume, and end with the student narrating their own transformation. Do this, and the written student record fills itself with things worth writing.

The point in designing a presentation event is to treat it as an assessment scene, not a showcase. Concretely, hand the rubric's perspectives to the audience too, and steer the questions toward the process. At an event where students are asked not "what was your conclusion?" but "where did you get stuck, and how did you get through it?", they have no choice but to narrate the process. A product outsourced to AI will always run aground in that Q&A. The guideline's suggestion of oral presentation opportunities³ earns its keep here as well.

One more thing: a portfolio that binds the process records becomes the student's own asset in admissions, such as comprehensive selection. Records created for assessment become the student's weapon as they are. Once that loop exists, record-keeping stops being drudgery imposed from above. And the timing matters: with the N-E.X.T. High School Vision, MEXT has placed inquiry-centered learning at the axis of high school reform, and I expect the design of assessment that underwrites the quality of inquiry to become a differentiator for schools over the next several years. I laid out the full picture in the guide to the N-E.X.T. High School Vision.

At TIMEWELL, through WARP for Schools, our program for schools and educational institutions, we help design inquiry-based learning that uses generative AI and build classes where AI chat logs are kept as records of learning. We have been involved in developing more than 500 people and have worked on a partnership project with the Tokyo Metropolitan Government (WARP ENTRE). If you are rethinking your inquiry program down to the assessment design, feel free to reach out — an exchange of notes is a fine place to start.

Summary

• Inquiry assessment is hard by structure. Each school defines the goals, content, and perspectives, and the student record is written prose rather than a number — so relief comes to the schools that design their own yardstick.
• The foundation is the three perspectives. "Attitude of proactive engagement" in particular is defined as the two sides of persistent effort and self-regulation of learning, not an impression to be graded.
• A rubric takes five steps: derive criteria from the stated purposes, narrow the scenes, write descriptors that bring actions to mind, share with students, and revise through use.
• In the generative AI era, building process records — chat logs, reflection sheets, fact-check records — into assessment deters outsourcing to AI and improves assessment validity at once.
• Presentations and portfolios close the loop. Hand the perspectives to the audience and point the questions at the process, and the event itself becomes your best assessment scene.

Assessment, pushed to its root, is a school's declaration of what kind of student growth it values. The hours spent debating rubric descriptors in the staff room are also hours spent sharpening the whole inquiry curriculum. Start small: half a page of A4, one rubric, for the "task setting" scene of your next unit.

---

References

Related articles

• 50 AI Inquiry-Learning Themes: A Ready-to-Use Collection for Teachers
• High School Inquiry-Based Learning × Generative AI: Practical Steps to Empower Students
• What Is the N-E.X.T. High School Vision? A Clear Guide to MEXT's High School Reform
• What Is AI for Science? How AI Is Changing Scientific Research, and Japan's National Strategy

Footnotes

1. MEXT, "Subjects Common to All Departments and Standard Credits" (2018 revision) ↩

2. MEXT, "Notification on the Improvement of Student Learning Assessment and Student Records in Elementary, Lower Secondary, Upper Secondary and Special Needs Schools" (No. 30-1845, March 29, 2019) ↩ ↩² ↩³ ↩⁴ ↩⁵

3. MEXT, Elementary and Secondary Education Bureau, "Guideline on the Use of Generative AI in Primary and Secondary Education (Ver. 2.0)" (December 26, 2024) ↩ ↩² ↩³ ↩⁴

4. National Institute for Educational Policy Research, Curriculum Research Center, "Reference Materials on Learning Assessment for the Integration of Instruction and Assessment: Upper Secondary School, Period for Inquiry-Based Cross-Disciplinary Study" (August 2021) ↩ ↩² ↩³ ↩⁴ ↩⁵

5. MEXT, "Comparison Table of the Courses of Study for Upper Secondary Schools: Period for Inquiry-Based Cross-Disciplinary Study" (2018) ↩

6. Central Council for Education, Elementary and Secondary Education Subcommittee, Curriculum Committee, "Report on the State of Student Learning Assessment" (January 21, 2019) ↩ ↩²

7. 48th Aomori Prefectural Upper Secondary School Education Research Association, Career Guidance Division Research Conference, keynote lecture transcript, "Entrepreneurship Education and Career Education" (FY2024 research bulletin) ↩