Using AI to Teach Scientific Inquiry in Grades 6-8

AI can support scientific inquiry in grades 6-8 by generating differentiated lab procedures, explaining the reasoning behind an experimental step, and translating raw data into different graph types for analysis. It must never predict or generate what an experiment's results "would" show, because collecting real data from real observation is the entire point of the practice, not a formality wrapped around a conclusion. The Next Generation Science Standards frame inquiry as a set of Science and Engineering Practices, not a body of facts to memorize.

Quick Answer: Use AI to generate leveled lab procedures, explain the purpose behind each experimental step, and help translate collected data into charts and graphs — while treating actual data collection as non-negotiable. An AI-generated "expected result" used in place of a real observation defeats the purpose of running the experiment at all.

Why Scientific Inquiry Is a Practice, Not a Body of Facts

For decades, many classrooms taught "the scientific method" as a rigid, memorizable sequence — hypothesis, procedure, results, conclusion — applied the same way to every lab regardless of what was actually being investigated. The National Academies of Sciences, Engineering, and Medicine's Framework for K-12 Science Education (2012), which the NGSS were built from, replaced that single linear method with eight Science and Engineering Practices: asking questions, developing models, planning investigations, analyzing data, using mathematics, constructing explanations, engaging in argument from evidence, and communicating information.

A rigid method treats science as a recipe to follow correctly; a practices-based approach treats it as a set of habits scientists actually use, in different orders, depending on the question.

That shift matters more than it sounds like a standards technicality — it changes what a "good" lab lesson looks like on any given day.

Middle school is where this shift is hardest to execute and most valuable to get right. Elementary science is often observation-heavy and low-stakes; high school science, where rigorous, frequently reverts to procedure-following under time pressure from content-coverage demands. Grades 6-8 sit in the gap where genuinely open-ended, phenomena-driven inquiry is both developmentally appropriate and still has room in the schedule.

Two lab styles illustrate the difference concretely:

• A "cookbook" lab tells students exactly what to do and what they should find — the outcome is predetermined before the lab starts, and the "investigation" is really just following instructions.
• A phenomena-based inquiry lab starts with something genuinely observable and puzzling, and lets the investigation design follow from students' own questions about it.

Neither style is inherently wrong for every lesson — a cookbook lab still has value for teaching a specific technique safely. The problem is a curriculum that's entirely cookbook, which trains procedure-following instead of the actual practices NGSS names.

Where This Sits in the Standards

NGSS is built on three-dimensional learning: Science and Engineering Practices, Crosscutting Concepts (patterns, cause and effect, scale), and Disciplinary Core Ideas (the actual content knowledge). A lesson that only covers content — skipping practices and crosscutting concepts — is only doing one-third of what the standards call for.

Anchoring phenomena is the NGSS instructional design term for a real, observable event or question that drives a full unit — why does a can of soda left in a hot car bulge, why do some objects float and others sink. The National Science Teaching Association (NSTA) has consistently promoted phenomena-based instruction as more engaging and more standards-aligned than a traditional topic-by-topic content march.

That framing matters for where AI fits. A good anchoring phenomenon needs to clear a few bars:

• Genuinely observable — students can actually see or experience it, not just read a description of it.
• Genuinely puzzling — it should raise a real question, not just illustrate a fact already explained.
• Accessible with real equipment and time — a phenomenon that requires specialized lab equipment a school doesn't have isn't usable, however scientifically interesting it is.

AI can help draft candidate phenomena and the discussion questions around them, but a teacher still has to check all three bars before committing a unit to one.

What AI Tools Can Actually Do for Inquiry-Based Science

Used well, AI's role is to remove friction around the practices that involve writing and explaining, while leaving the actual empirical work — running the investigation, collecting the data — entirely to students.

Generating Differentiated Lab Procedures and Safety Checklists

A tool like EduGenius can generate a leveled lab procedure from a class profile — more scaffolded step-by-step instructions for students new to lab work, a more open-ended procedure for students ready to design more of the investigation themselves — plus a safety checklist matched to the specific materials and hazards involved.

Explaining the "Why" Behind a Procedure Step

Students often follow lab steps mechanically without understanding why a specific step matters — why a control group is necessary, why a variable has to be isolated. AI-generated plain-language explanations of the reasoning behind a procedure can turn a step a student was just executing into one they actually understand, without changing what the student physically does in the lab.

Translating Raw Data Into Different Graph Types

A student who collects a table of raw numbers often doesn't yet know which graph type best reveals the pattern in that specific data. AI tools can suggest and generate a bar graph, line graph, or scatter plot from the same dataset, letting students compare representations and discuss which one actually communicates the finding most clearly — a skill NGSS's "analyzing data" and "communicating information" practices both call for directly.

Generating Open-Ended Follow-Up Questions

The best inquiry labs don't end when the data is collected — they end with a new question the results raise. AI can generate a set of follow-up "what would happen if" prompts tied to a class's actual findings, extending a single investigation into a genuine next question instead of stopping at "we found X."

This matters because generating a good follow-up question on the spot, for 28 different sets of results, is a lot to ask of one teacher circulating a room — exactly the kind of repetitive-but-necessary drafting work AI handles well.

The One Line AI Must Never Cross: Predicting Results Instead of Collecting Them

This is the single most important caution in this entire piece. Asking an AI tool "what would this experiment show" and using its answer instead of running the actual investigation isn't a shortcut — it replaces the empirical practice itself with a guess dressed up as a result.

A generative AI model has no way to know what a specific class's actual materials, conditions, and measurement error will produce on a given day. Its answer, however confident and plausible-sounding, is not data — and presenting it as data to students teaches exactly the wrong lesson about where scientific knowledge comes from.

That last row deserves its own note: real experiments sometimes produce messy, unexpected, or "wrong" results, and that's genuinely useful scientific information, not a failure to hide. Grading against an AI-generated expected outcome risks teaching students to adjust their data to match a guess instead of reporting what they actually observed.

A Classroom Walkthrough: A Phenomena-Based Inquiry Unit

Say you teach seventh-grade physical science and want to anchor a unit on heat transfer with a genuinely observable phenomenon — condensation forming on the outside of a cold drink container on a warm day.

• Before the investigation: you could use an AI tool to generate a leveled lab procedure for a related, safe classroom investigation (measuring temperature change across different container materials), plus a set of question prompts to help students articulate what they're curious about after observing the phenomenon.
• During the investigation: students collect their own temperature and time data — this step has no AI involvement at all, since the entire point is real measurement under real conditions.
• After the investigation: an AI tool can help students generate a few different graph types from their collected data, and draft explanatory language they then revise using their own findings and reasoning, not a generated conclusion.

At no point does the AI tool tell students what they should have found — it supports the practices around the data, never the data itself. If two lab groups get noticeably different temperature curves from the same setup, that's a discussion about measurement consistency and sources of error, not a signal that one group's real data was wrong.

A Practical Framework for a Scientific Inquiry Unit With AI

Say you're building a two-week inquiry unit for a mixed sixth-grade class, anchored by a genuinely observable phenomenon. Here's a sequence that keeps the empirical work with the students.

1. Start with a real, observable phenomenon, not a AI-generated hypothetical one — anchoring phenomena work because students can actually see them, not because they're described well.
2. Generate differentiated procedures and safety materials with AI, reviewed by you for accuracy and available equipment before distribution.
3. Collect real data with no AI involvement. This step is non-negotiable regardless of how tempting a shortcut might be under time pressure.
4. Use AI to help visualize and represent the collected data, comparing graph types to find which one best communicates the actual finding.
5. Construct explanations using the real data, with AI available for language and structure feedback — never for supplying the conclusion itself.

Assessing Inquiry Skill, Not Just Lab Report Formatting

A polished, well-formatted lab report can still reflect weak inquiry skill — copied procedural language, data that was quietly adjusted to look cleaner, a conclusion that doesn't actually follow from what was measured. NGSS's practices-based approach means assessment should target the practices directly, not just the final report's presentation.

• Ask students to defend an unexpected result. A student who can explain why their data came out differently than predicted, using sound reasoning about sources of error, has demonstrated real understanding — possibly more than one whose data happened to match expectations.
• Check the data-to-conclusion link, not just the conclusion. Does the stated conclusion actually follow from the specific numbers collected, or could it have been written before the lab even ran?
• Assess practices across a unit, not just one report. Asking a good question, planning an investigation, and arguing from evidence are separate skills that don't all show up equally in a single lab write-up.

AI-generated rubric language tied to specific NGSS practices can help build assessment criteria that target the practice being taught, rather than defaulting to a generic five-section lab report format that rewards formatting over actual scientific reasoning.

Comparing Tools for Middle School Scientific Inquiry

No single platform covers simulation, real data collection, and worksheet generation equally well. The table below compares what middle school science teachers most often reach for.

A workable setup pairs real data-collection hardware like Vernier sensors (or simply a thermometer and stopwatch) for the actual investigation with a simulation tool like PhET for concepts too costly or dangerous to run physically, then uses a generator like EduGenius for the leveled procedures and rubrics built around both.