NSF-SBIR Funded Research

Evidence-based AI for education

Two years of funded R&D, 100+ interviews, and validated results—not another ChatGPT wrapper.

The assessment crisis is real

These aren't our numbers. They're from peer-reviewed research.

92%

of students use AI in studies

HEPI Survey 2025

88%

use generative AI for assessments

HEPI Survey 2025up from 53% in 2024

89%

use ChatGPT for homework

Forbes/multiple 2024-25

29%

have had ChatGPT write entire essays

Intelligent.com 2024

AI detection doesn't work

50%

false positive rate in real-world testing

Washington Post

Vanderbilt disabled Turnitin AI detection

Vanderbilt 2023

Higher false positives for ELL & neurodivergent students

Stanford HAI

"The epistemological crisis in higher education caused by AI-generated student work."

Chris Duke, Vice Provost, San Jacinto College

Proven in blind tournament conditions

Phase I results from double-blind evaluation against humans and frontier LLMs.

SourceOverall ScoreEvidence QualityCitation Accuracy
DebaterHub81.786.776.2%
Human Experts70.156.98.7%
Zero-shot LLMsGPT-4, Claude, Gemini, Grok50.627.10%
11.6

points better than expert human debaters

31.1

points better than frontier LLMs

76% vs 0%

citation accuracy gap vs ChatGPT

IBM ultimately conceded its debate systems were incapable of engaging in expert human-level debate. — Slonim et al., 2021

Dialogical Proficiency

A validated rubric for measuring what matters in argumentation.

Research Foundation

Built on Dr. Robin Alexander's dialogic teaching research, which showed:

English progress+2 months
Science progress+2 months
Mathematics progress+1 month
Disadvantaged learnersComparable gains

Alexander, 2018; Jay et al., 2017 (RCT)

Scoring Dimensions

Evidence Use

Quality and relevance of supporting materials

Reasoning Quality

Logical structure and warrant strength

Engagement with Counterarguments

Responsiveness to opposing views

Synthesis

Integration of multiple perspectives into coherent position

"AI, paradoxically, enables us to assess what truly matters for democratic education: dialogical proficiency—the ability to engage in structured, reasoned discourse across difference."

Hines, 2025

100+ interviews. Here's what we heard.

NSF I-Corps customer discovery with educators, administrators, and students.

Data Ownership Concerns

"ACU signed a deal with OpenAI and administrators are not happy about some of the terms. The most frustrating and urgent need we could fill is the lack of data ownership, citing intellectual property concerns."

— Anne Marie Todd, Professor & Dean, San Jose State University

"Pushback against the OpenAI partnership, with data wall being a primary sticking point. Usage limits were also woefully limiting."

— Shikar Sethi, Entrepreneur, AssessPrep

Assessment Crisis

"Started as an Anti-AI person but is now going to recommend entire First Year Writing program participate in the Pilot... Our focus on dialogical assessment of AI-Proof skills was exciting."

— Jacqueline Foertsch, Professor (First Year Writing)

"AI implementation in classrooms raises concerns about academic integrity and student skill development."

— I-Corps Interview Notes

Value of Dialogical Assessment

"Loved the idea of taking rubric-based grading out of hands of instructors so they could focus on leading class discussions and engaging students on developing their ideas and their voice."

— Andrew Leong, Professor of English, UC Berkeley

"Assessment is a key differentiator we need to lean into."

— Michael Street, Lead Editor, Great Minds PBC

Teacher Pain Points

"Teachers are leaving the profession because they don't get paid and don't get respect. Administrators are worried about saving money so are also firing teachers. Those who remain are drowning in grading and paperwork and can't focus on developing skills and ideas with students."

— I-Corps Interview (Assessment/EdTech expert)

"Biggest challenge is constant race to close achievement gaps while also having to wear numerous hats at the same time. Biggest improvement from AI is increased time to focus on things that matter."

— Bayleigh Witman, K-12 Educator, Llano ISD

LMS Integration Requirements

"Ease of integration into existing LMS is extremely important."

— I-Corps Interview (K-12 Tech Director)

"Need for a value proposition document explaining the concept and benefits... Interest in integrating the solution into existing systems like Blackboard or Canvas."

— Chris Duke, Vice Provost, San Jacinto College

Market Differentiation

"The vast majority of startups at the tradeshow are offering similar and undifferentiated products that all looked to be no more than AI Wrappers. DebaterHub was the only discussion that shared a 'unique vision.'"

— Anjali Tiwary, CEO, Indian Debating League (90,000 students)

"Strong customer market potential as B2C play in India. There's investable value in creating new assessment signals."

— Atul Thakkar, Director, Anand Rathi Investment Banking

Pilot institutions

Research partnerships validating the approach in real classrooms.

University of North Texas

Entire FYW curriculum (thousands/semester)

Champion: Marshall Armintor, Brian Lain

STTR Partner

San Jacinto College

Community college pilot

Champion: Chris Duke

Letter of Interest

Southern Methodist University

Honors seminars

Letter of Support

University of Mary Washington

Gen Ed Communications

Champion: Anand Rao

Co-PI

Illinois State University

Summer bridge program

Champion: Byron Craig

Exploring

UC Berkeley

FYW/English

Champion: Andrew Leong

Interested

International Partners

Indian Debating League

Anjali Tiwary

Investment interest, channel partner

NAUDL

In discussion

OratorLab

Taiwan/Asia

Partnership

Publications & Presentations

Augmented Debate-Centered Instruction

2024

Responsible AI integration in education through debate-based learning.

AAAI-24

AI Pluralism

John Hines

2025

Theoretical foundation for pluralistic AI systems that preserve belief diversity.

Palgrave Macmillan

NSF SBIR Phase I Final Report

2024

Technical validation of AI debate system against human experts and LLMs.

National Science Foundation

NSF STTR Phase II Application (under review)

2025~$1.5M (pending)

Proposal for randomized control trials and university pilots for dialogical assessment.

National Science Foundation

Technical differentiation

What makes our AI different from ChatGPT wrappers.

Proprietary debate datasets

85,000+ evidence cards with validated citations and source material.

Dialogical assessment pipeline

Multi-dimensional scoring validated against expert human judges.

Pluralistic worldview system

50+ coherent perspectives that maintain internal consistency during debates.

Sub-500ms voice latency

Real-time conversation flow without awkward pauses.

LTI 1.3 native integration

Deep Canvas integration with automatic grade passback.

Want to pilot with us?

We're looking for research partners to validate dialogical assessment in more contexts.

Color scheme toggle