50 Inquiry-Learning Themes for High School (AI & Frontier Science): A Ready-to-Use List of Questions

This is Hamamoto from TIMEWELL.

The single most common request I hear from teachers about the period for inquiry-based study is help with setting themes. "I want students to do inquiry around AI or frontier science, but they don't know what to ask." "I offer an interesting-looking theme, and they just ask AI and call it done." This article is for exactly those teachers. It collects 50 inquiry themes built around AI and frontier science, organized into a table by field: life science, math and physics, robotics, social issues, community, and ethics. Simply listing themes does not make a lesson, so I have also written out how good and bad questions differ, how to frame questions in each field, and how to keep students from handing everything off to AI, all in a form you can drop straight into your design.

Let me lay out the three ideas underlying this article.

• An inquiry theme should not be a question whose answer pops out when you look it up. It should be verifiable with primary sources or data, and connected to the student's own life or community.
• Frontier science such as AlphaFold3, genome-analysis AI, AI at the Math Olympiad, and Physical AI is published as primary information by developers and public agencies, so a familiar entry point makes it usable even for high schoolers.
• AI can be a partner at every stage of inquiry, but drawing the line in advance between what AI handles and what the student handles prevents the hand-off.

The period for inquiry-based study is a required subject of three to six standard credits under the 2018 Courses of Study for high schools¹. Its stated aim writes the inquiry process directly into the text: "find questions from the relationship between real society and real life and oneself, set your own task, gather information, organize and analyze it, and summarize and express it"¹. Choosing a theme is precisely the design of that "finding a question" stage. Get it wrong, and the later information-gathering and analysis spin their wheels. If you first want to gauge how far you personally could bring AI into your class, taking our AI literacy check before you start designing themes is one reasonable place to begin.

Where Good and Bad Inquiry Themes Part Ways

Before the 50 themes, the most important point. Hand students a theme list and they gravitate toward whatever sounds impressive. But the success of an inquiry is decided not by how flashy the topic is, but by the quality of the question. When I refine themes together with teachers, I always use two yardsticks: verifiability and personal relevance.

Verifiability is whether the question can be checked with data or primary sources. "Is AI good or bad for humanity" looks grand, but it is completely beyond what a high schooler can verify in three months. It ends with the student having AI write a plausible-sounding generality. By contrast, "Ask 100 students about our school's AI-use rules, classify the reasons for and against, and propose a line everyone can accept" can be verified with the tangible data of a survey.

Personal relevance is whether the question connects to the student's own way of living or to the place they live. The aim in the Courses of Study, "while considering one's own way of living," points to exactly this¹. A question about someone else moves no one, no matter how much you research it. The moment it is pulled toward one's own path or a local issue, students lean in on their own.

Using these two yardsticks, common bad questions can be refined into good ones as follows.

Refining lens	Bad question	How to refine it
Verifiability	Is AI good or bad for humanity	Ask 100 students about our AI-use rules, classify the reasons, and propose a line
Primary sources	What is AlphaFold	Read the official explanation, pick a drug used at a local hospital, and hypothesize the shortening of its development time
Personal relevance	Will AI take away jobs	Predict whether the job you want will still exist in ten years, backed by data you collect
Question size	About AI and society	Sort school cleaning tasks into those a real-world AI can handle and those it cannot
Taking a position	About AI ethics	Gather both sides on whether creators of training-data works deserve compensation, then state your own position

What the right column shares is a structure in which the answer is not fixed to one, and the student cannot proceed without going somewhere or asking someone. That structure is the very thing that makes a question impossible to hand off to AI. All 50 themes below are built on this idea. The numbers run straight through, so pick from them to match your grade level and your class's interests.

Life Science x AI Inquiry Themes

Life science is the field where AI has most dramatically changed how research is done. The 2024 Nobel Prize in Chemistry went to Demis Hassabis and John Jumper, who predicted protein structures with AI, and to David Baker, who designed proteins by computation². Determining a protein's shape, which used to take years, can now be predicted with high accuracy in minutes. AlphaFold3, released in May 2024, can predict not only proteins but also DNA, RNA, and interactions with drug molecules, according to its developer³. On the genome side, Evo 2, trained on more than 128,000 genomes across the three domains of life, reportedly judged whether variants in the breast-cancer-related gene BRCA1 are benign or pathogenic with over 90% accuracy⁴.

#	Inquiry question	Entry point for verification / primary sources
1	When AI can predict a protein's shape, what changes in developing familiar drugs?	Pick one drug used at a local hospital or pharmacy, look up its years to approval, and hypothesize
2	If genome AI reveals disease-prone variants, how should health checkups change?	Survey 100 students on the benefits and worries, and classify the responses
3	Can you explain to a middle schooler why the 2024 Chemistry Nobel went to AI protein work?	Read the official award rationale and retranslate it into plain language
4	If AI can design new proteins, can we make foods less likely to trigger allergies?	Gather familiar allergy labels and consider feasibility
5	With AI proposing drug candidates, why can't human clinical trials be skipped?	Research the risk of skipping them, and draw a line with what AI may do
6	What do you see when you visualize a textbook protein with a free prediction tool?	Actually display the structure and compare it with the textbook figure
7	What accuracy percentage should AI reach before it is used in clinical settings?	Set your own passing line, give reasons, and ask a teacher or medical professional
8	How many years could genome AI cut off improving a local crop variety?	Interview a farmer or agricultural co-op and compare with conventional breeding
9	When AI can write the blueprint of life, who should draw the line on organisms we must not make?	Research domestic and overseas rules and guidelines, and write your own proposal

The trick for framing questions here is to always land the cutting-edge result on a familiar object. Stopping at "AlphaFold is amazing" produces a book report. Having the student pick one object they can verify in reality, such as a drug at a local hospital, a food label at home, or a local farm product, turns an abstract topic into inquiry at once.

To stop the hand-off, always assign students the one step of checking whether AI's explanation is true against primary sources. Both AlphaFold3 and Evo 2 publish their performance on official blogs and in papers. Asking AI "what is AlphaFold" is research, not inquiry. Make it the student's job to confirm the official figures with their own eyes and then use those figures to hypothesize about a local subject, and AI stays a research partner.

Math and Physics x AI Inquiry Themes

Because they are not memorization subjects, math and physics tend to be avoided as inquiry themes. Yet it is precisely here that AI's progress is producing exciting questions. On FrontierMath, a benchmark of the world's hardest math problems, leading AIs at first solved fewer than 2%, but within just a few months the success rate rose past 20%⁵. In July 2025, a Google DeepMind model scored 35 out of 42 at the International Mathematical Olympiad and was officially recognized by the organizers as meeting the gold-medal standard⁶. The same company's AlphaEvolve has begun to overwrite math's best-known results, updating a record on a geometry problem studied for over 300 years⁷.

#	Inquiry question	Entry point for verification / primary sources
10	AI's accuracy on the hardest math problems surged in months. Will it continue or plateau?	Graph the published success rates and predict the future yourself
11	For AI's Olympiad gold, what separates "official recognition" from "self-reported"?	Compare the developers' announcements and distinguish whether it was certified
12	Now that AI updates math records, how does the human mathematician's role change?	Interview a math teacher or a university student and organize the views
13	If AI optimizes the school timetable or club shifts, is it better than the human draft?	Decide criteria first, then score the AI draft and the handmade draft
14	Can a high schooler check a proof written by AI?	Pick one problem and trace each line of the proof yourself
15	How far is AI already used in weather and earthquake prediction?	Research public-agency announcements and log hits and misses for a month
16	When AI analyzes physics data, what becomes easier and what is lost?	Compare hand calculation and AI analysis on pendulum or free-fall measurements
17	Does AI get more accurate the longer it thinks?	Try the same problem under varied conditions and log time versus accuracy

The trick for framing questions is to pull AI's results down from "news" to "experiment." Researching the gold medal itself ends in a summary of an article. Instead, have the student actually pit AI against a human draft on a familiar optimization problem, or check an AI proof line by line, and math becomes an object of inquiry.

To stop the hand-off, keep the initiative for verification in the student's hands. Do not end at having AI produce an answer; always include a step where the student confirms whether it is correct. For timetable optimization, the student judges the AI's draft against the criteria rather than swallowing it whole. Inquiry collapses the moment the answer-checking is handed to AI, so make sure the judgment stays human.

Robotics and Physical AI Inquiry Themes

After AI that generates text and images in digital space comes Physical AI that perceives, judges, and moves a body in the real world. NVIDIA's CEO Jensen Huang emphasized the concept in his CES keynote in January 2025, saying the turning point for general-purpose robots is near⁸. The company's Isaac GR00T N1, released in March 2025, is described as the world's first open foundation model for humanoids, working with a two-part design of a fast reflexive system and a slower planning system⁸. Automating experiments is advancing too. A "mobile robotic chemist" built by a University of Liverpool team ran autonomously for eight days, carried out 688 experiments, and found a photocatalyst combination about six times more active than the initial recipe⁹.

#	Inquiry question	Entry point for verification / primary sources
18	How far can a real-world AI take over school cleaning or serving meals?	List campus tasks and sort them into those it can and cannot handle
19	As robot "brain foundations" become shared, how do local factory automation costs change?	Interview local manufacturers about the barriers to adoption
20	The 688 experiments a self-driving lab ran in eight days would take a human how long?	Estimate the time per experiment and calculate the value of leaving it to AI
21	When a self-driving car causes an accident, who should be responsible?	Research the debate and propose conditions for introducing it in your town
22	How do older people themselves feel about care or watch-over robots?	Interview a local facility with permission
23	Viewing old campus devices through a "Physical AI" lens, where could you make them smarter?	Observe automatic doors and ticket machines and diagram where to add sensors and judgment
24	How far does robot automation solve a region's labor shortage in farming?	Gather trends in the number of farmers and show both the gains and the limits
25	Do humanoid robots really need to be humanoid?	Design the optimal shape per task and compare it with the humanoid form

The trick for framing questions is to tie robots not to objects of admiration but to specific tasks on campus or in the community. Watching humanoid-robot videos does not make inquiry. List real tasks such as school cleaning, a local factory, or farm work, and have students consider how to add AI and sensors there, and you get robotics inquiry cut to their own size.

To stop the hand-off, make on-site observation and interviews mandatory. In robotics you only understand the problem by seeing the real thing. A record from visiting a factory or a care facility and talking with people is something AI simply cannot generate. Build on the primary information the student gathered, and let AI help only with organizing the material. Keep that order and AI stays a set of training wheels.

Social Issues x AI and Community x AI Inquiry Themes

This is the field easiest to make personal. As generative AI spreads, the content of jobs shifts, misinformation grows, and a gap opens between those who can and cannot use AI. Because these issues connect directly to students' own paths and lives, they can be dug into deeply once the question is designed. Turn to the community and you find inconveniences you would like AI to solve everywhere: tourism, traditional industry, transport, disaster prevention, farming. The national government, too, is pushing to implement AI in society, for example by spreading small research grants for using AI across all fields to a scale of some 1,000 projects¹⁰.

First, social-issue themes.

#	Inquiry question	Entry point for verification / primary sources
26	With generative AI, what jobs disappear and what jobs grow?	Pick a job you want and predict, with evidence, whether it survives in ten years
27	Can classmates tell AI-written text from human-written text?	Design and run a blind test and produce the accuracy rate
28	How should our school's AI-use rules be set?	Research the current state and draft a policy students and teachers both accept
29	Is AI-based grading or admissions use fair?	Research past unfair cases and propose conditions for fairness
30	Should creators of works used in training data be compensated?	Gather both sides and state your own position with reasons
31	With AI mass-producing misinformation, how can a high schooler spot what is true?	Build a checklist for telling real from fake and test it on actual fakes
32	How should we use a human confidant versus an AI confidant differently?	Survey actual usage and write guidance on when to use which
33	Will a new gap open between those who can and cannot use AI?	Verify with data on home internet access and device ownership
34	How much electricity and water does a day of your AI use consume?	Log your usage count and estimate consumption from public data

Next, community themes. With your hometown as the stage, both interviewees and verification data are close at hand.

#	Inquiry question	Entry point for verification / primary sources
35	If you build multilingual AI tourist guidance for your area, are visitors really helped?	Make a prototype, have visitors and locals use it, and collect feedback
36	Can AI record and pass on the skills of a traditional industry?	Interview an artisan and separate what AI can take on from what only a person can pass down
37	Is AI on-demand transport usable as mobility in a depopulated area?	Research your town's bus routes and ridership and draft an introduction plan
38	How can AI prediction help local disaster prevention?	Overlay the hazard map with past disaster records to identify dangerous spots
39	How can AI help use empty storefronts in the shopping district?	Hypothesize foot traffic and sales and run an opening simulation
40	Is there meaning in using AI to record and preserve a vanishing dialect?	Interview the grandparent generation and find a way to record and keep the words
41	How far can AI and sensors make local farming more efficient?	Shadow one farm and lay out the problems and effects
42	Which "inconvenience you want AI to solve" in your town has the biggest impact?	Ask 100 residents, rank them, and dig into the top one

The trick for framing questions is to start not from "what can AI do" but from "what around me do I want to solve." Make technology the subject and you chase trends. Make the trouble the subject and verify whether AI is a sound solution to it, and you get inquiry infused with the student's own sense of reality. Empty storefronts, the commuter bus, a grandparent's dialect. The best themes often sleep within a few kilometers.

To stop the hand-off, always build in a mechanism for the student to go collect primary data themselves. Surveys, interviews, and usability tests of prototypes cannot be replaced by AI. For spotting misinformation, do not have AI write generalities; have the student actually collect fakes and judge them with a checklist. With data gathered by their own hands, AI serves only as the organizer and the conclusion belongs to the student. As an entry point, reading how to embed generative AI in high school inquiry alongside this gives you the full picture of task design that prevents the hand-off.

Ethics and AI Literacy Inquiry Themes

Last is the field most like inquiry, precisely because the answer is not fixed to one. MEXT's N-E.X.T. High School Vision places "abilities AI cannot replace" at the core of future high school education, stating clearly that the ability to frame one's own questions should be valued over the sheer volume of memorized knowledge¹¹. Thinking about how to live alongside AI is itself at the center of the learning this vision calls for. Here I list questions that have students draw their own lines with their own heads.

#	Inquiry question	Entry point for verification / how to deepen it
43	Is having AI do your homework "cheating"?	Define your own boundary for what is and is not cheating, and debate it
44	How do the scenes where you can trust AI and where you should doubt it divide?	Gather cases of misinformation and hallucination and classify trust versus caution
45	Is the "fear of AI taking jobs" actually correct?	Compare with how jobs changed under past innovations like the car and the PC
46	Should technology that reads human emotion be brought into schools?	Organize the trade-off with privacy in your own words
47	To keep AI from making discriminatory judgments, what can developers and users each do?	Make a concrete action list and examine whether it is achievable
48	Is there meaning in humans learning "things you can just ask AI"?	Using times tables and kanji memorization, state your view with reasons
49	What are the risks of entering personal information into generative AI?	Actually read the terms of service and set your own rule for what to enter
50	Ten years from now, what kind of person do you want to be with AI?	Interview an admired adult or professional and declare one ability you want to value

The trick for framing questions is not to end with the large container of "AI ethics," but to have students draw their own lines. Ethics is not commentary; it is judgment. Prepare concrete scenes such as homework, personal information, and emotion recognition, and have the student decide "how far would I allow it," and abstraction turns into their own judgment. Question 50 works for a wrap-up or a pre-graduation reflection.

To stop the hand-off, tell students that ethics is the very domain not to let AI answer. Ask AI to "discuss AI ethics" and back comes a harmless generality. That is not the student's opinion. Gathering the material for both sides and organizing cases can go to AI, but the final position is stated by the student in their own words. Keep that, and ethics inquiry becomes the student's living conclusion. For broader context, see what the "abilities AI cannot replace" are and the introductory guide to AI for Science, which lays out how AI is changing scientific research.

Instructional Design That Prevents the Hand-Off, and a Wrap-Up

I have laid out 50 themes, but the themes themselves are not what I most want to convey. It is that any theme can become inquiry or become busywork, depending on how the question is designed and where you draw the line with AI. At TIMEWELL we provide hands-on AI support for companies, and I have taught AI use to several hundred adults. What that drove home is that the people who use AI best are the ones who can draw the line between the work they may leave to AI and the work they should not. That sense of where to draw the line settles in far better when you build it once in high school inquiry than when you try to acquire it as an adult.

In the classroom, I recommend having students decide, on a single sheet at the start of the inquiry, "the part we leave to AI and the part we handle ourselves." Research, rough drafts, and organizing data may go to AI. Setting the question, checking primary sources, local interviews and surveys, and stating the final position stay with the student. Put that line into words in advance, and when the temptation to "just ask AI here" arrives midway, the team can return to their own rule. The hand-off is not the student's laziness; it is a problem on the adults' side for not designing the line, or so I believe.

At TIMEWELL, with this thinking, we offer a WARP program for schools and educational institutions. From designing inquiry themes to walking alongside students until they build a product with AI through to the end, we assemble it around each school's situation. We have supported the development of more than 500 people through our programs for working adults, and we have run WARP ENTRE as an agreement project with the Tokyo Metropolitan Government. If you are a teacher wrestling with how to connect inquiry and AI, or unsure how to bring frontier-science themes into class, please reach out, even just to compare notes.

Finally, one watchword for when you are stuck choosing a theme. "Can the student proceed without verifying on site or asking someone?" If you can answer no, the students cannot hand it off to AI. As you look over the 50-row table, picture the face of the student in front of you and pick just one question.

---

References

Related Articles

• High School Inquiry-Based Learning x Generative AI: Practical Steps to Empower Students
• What Are the "Abilities AI Cannot Replace"? How AI Education in High Schools Is Changing
• What Is the N-E.X.T. High School Vision? A Clear Guide to MEXT's Grand Design for High School Reform
• What Is AI for Science? A Clear Guide to How AI Is Changing Scientific Research

Footnotes

1. MEXT, "High School Courses of Study (announced 2018)" and materials on the period for inquiry-based study and standard credits (accessed 2026-07-18) ↩ ↩² ↩³

2. Google DeepMind, "Demis Hassabis & John Jumper awarded Nobel Prize in Chemistry" (October 9, 2024) ↩

3. Google, "AlphaFold 3 predicts the structure and interactions of all of life's molecules" (May 8, 2024) ↩

4. Arc Institute, "Evo 2" (February 19, 2025) ↩

5. Epoch AI, "FrontierMath" benchmark overview ↩

6. Google DeepMind, "Advanced version of Gemini with Deep Think officially achieves gold-medal standard at the IMO" (July 21, 2025) ↩

7. Google DeepMind, "AlphaEvolve: a Gemini-powered coding agent for designing advanced algorithms" (May 14, 2025) ↩

8. NVIDIA, "Isaac GR00T N1: Open humanoid robot foundation model" (March 18, 2025) ↩ ↩²

9. Chemical Reviews, "Self-Driving Laboratories for Chemistry and Materials Science" (original source: Burger et al., Nature 2020) ↩

10. MEXT, "Program for Transforming Scientific Research through AI for Science (SPReAD)" (accessed 2026-07-18) ↩

11. MEXT, "Basic Policy on High School Education Reform (Grand Design): The N-E.X.T. High School Vision toward 2040" (February 13, 2026) ↩