Student-Led · Research-Driven · Impact-Focused

Teaching students to think better.
Measuring whether it works.

ScienceTech Foothold helps students move from having opinions to building arguments — with a clear claim, a specific reason, real evidence and a fair look at the other side.

See Our Impact → How We Measure It

Scroll

Students Online

Reached through social media and digital outreach

Students In Person

Reached through offline fieldwork and classroom teaching

0.0

Avg. Reasoning Gain

Absolute improvement across the classes evaluated

Student Team

Members working on the field implementation

Why We Exist

We all have opinions.
But can we support them?

In classrooms we kept seeing the same thing: students had strong views, but when asked why, the answer stopped at "because it's true" or "everyone knows that." A clear claim, a specific reason, relevant evidence and a moment of considering the opposite view were often missing.

The gap isn't a lack of intelligence. It is a skill gap — and skills can be taught, practiced and evaluated.

How It Started

ScienceTech Foothold began as a Class XII General Studies project at City Pride School, Nigdi Pradhikaran. Instead of submitting only a written report, the team ran a real classroom intervention: they observed the reasoning gap, designed a teaching module, collected baseline responses, taught the sessions, collected post-assessments and built a Machine Learning model, trained on IBM and ASAP AES argument quality scorer datasets, to compare the two consistently.

Image 2

Students During Session

Image 3

Team Photo

Founded by

Devang Shimpi

Student Lead · with a six-member student team

What We Teach

Five questions that turn an opinion into an argument.

Every session is built around one simple, repeatable structure. Students learn it, practice it on real prompts, and check their own writing against it.

Claim

What are you actually trying to prove? One clear, specific sentence.

Reason

Why do you believe it? The logical bridge from claim to evidence.

Evidence

What can actually support it — and how reliable is that source?

Other Side

What would someone who disagrees say? Address it honestly.

Conclusion

What follows once all the evidence is considered — without absolute language.

Think Like a Scientist

Question→ Observe→ Check Evidence→ Conclude→ Revise

The same evidence-first loop applies to viral messages, school debates, social media, everyday decisions and public claims. We teach students to run it on purpose.

Also covered in every module

Opinion vs FactEvidence QualityHidden AssumptionsOvergeneralisationFalse CauseBandwagon BiasConfirmation BiasSource ReliabilitySelf-Checking

From the Classroom to the Field

This wasn't a paper exercise.
We went into classrooms.

The student team delivered in-person reasoning sessions to school students, reaching 500+ students through offline fieldwork. The documented General Studies intervention covered Standards 6, 7 and 8 at several schools and communities with structured instruction and matched pre/post assessment.

Image 4

Classroom Teaching

Student-led instruction on Claim → Reason → Evidence.

Image 5

Pre-Assessment

Students writing baseline responses before any teaching.

Image 6

Post-Assessment

Matched post-assessment on comparable prompts.

Image 7

Group Photo

Students and the team after a session.

Methodology

We didn't just teach it.
We measured it.

Identify the gap

Students had opinions but lacked structure and evidence.

Design the module

Structured curriculum plus comparable assessment prompts.

Pre-assessment

Collect baseline written responses before teaching.

Instruction

Teach the framework, scientific thinking and reasoning traps.

Post-assessment

Matched responses, evaluated on the same scale.

Measuring Reasoning with AI

AI is not the intervention.
It is the measurement layer.

Grading hundreds of handwritten arguments by hand is slow and inconsistent. So the team developed a transformer-based NLP model that scores each written response on the same four dimensions — logical structure, reasoning depth, evidence quality and clarity — before and after the sessions.

The model was developed using academic argument-quality datasets, including the ASAP and IBM Argument Quality datasets, and deployed as the Argument Intelligence Analyzer. Its job is consistency: every student's pre- and post-response is judged by the same standard.

Student response→ AI analysis→ Reasoning score→ Pre vs post

Argument Intelligence Analyzer

Demo · Mock UI

Student response

"Schools should reduce exams because frequent high-stakes testing pushes students toward memorisation. While exams give useful feedback, combining them with projects gives a fuller picture of understanding."

Logical Structure84

Reasoning Depth79

Evidence Quality72

Clarity88

Overall Reasoning Score

82 / 100

Try the Argument Intelligence Analyzer →

Did It Work?

21.5%

Average absolute improvement in logical reasoning.

Students completed a pre-assessment before the sessions and a matched post-assessment after. Both sets were scored on the same reasoning scale using the same AI evaluation framework. Across the classes covered by the evaluated intervention, the average absolute gain was 21.5 percentage points.

Pre-assessmentBaseline

Post-assessment+21.5 pp

Cohort A

17.7pp

Cohort B

20.7pp

Average

21.5pp

Bars are illustrative of the gain, not raw score values. Documented matched-cohort gains shown.

What we did

Taught 500+ students in person

Structured reasoning sessions across Standards 6, 8 and 9, plus 1,000+ students reached online.

What we measured

Pre / post reasoning scores

Matched written responses scored on structure, depth, evidence and clarity by one consistent model.

What we found

+21.5 percentage points

Average absolute gain in logical reasoning across the evaluated classes.

What's next · future scope

Scale it, and test it harder

· More classrooms and schools
· Delayed assessments to test retention
· Refine the model with more annotated responses
· Extend to Science and History
· Open reasoning resources for any student

What We're Still Learning

We care about impact — and about whether our evidence supports our claims.

Short duration. The evaluated intervention ran over a relatively brief period.
Sample size. Larger and more diverse samples would strengthen the findings.
Retention. Long-term retention has not yet been tracked.
Model limits. AI evaluation can misjudge unusual responses and needs continued refinement.

Image 8

Devang Shimpi · Founder

The People Behind It

Devang Shimpi

Founder · Student Lead

Devang started ScienceTech Foothold with one question: can a short, structured intervention actually improve how students reason — and can we prove it? What began as a General Studies project became a documented classroom study, an AI-supported evaluation tool and an ongoing student-led initiative.

The six-member project team

Kanishka Sharma

Anay Telang

Aakruti Tulsiani

Laksh Vijayvargia

Harshwardhar Singh

Yashasvi Singh

Learning Resources

The modules we teach from.

Cover

Class 6

Think Smart: How to Build Strong Arguments

A first introduction to claim, reason and evidence, spotting thinking traps, and telling opinion from fact.

Explore →

Cover

Class 7

Class 7 · Critical Thinking

Reason Like a Scientist

Question → Observe → Check Evidence → Conclude, applied to everyday claims, source reliability and cognitive bias.

Explore →

Cover

Class 8

Class 8 · Critical Thinking

The Art of Strong Reasoning

Hidden assumptions, evidence quality, handling the other side, avoiding absolute language and self-checking your own argument.

Explore →

Get Involved

Help us build
better thinkers.

We believe reasoning is a skill — and skills can be taught, practiced and measured. Bring a session to your school, volunteer with the team, or ask us about the research.

✉

devangpshimpi@gmail.com

◉

Instagram

@sciencetech_foothold