RESEARCH & INNOVATION
Advancing the Future
of Learning
Learnology Labs is a premier academic research institute dedicated to rigorous, evidence-based inquiry. We bridge the gap between educational theory and transformative digital practice to shape the future of learning.
OUR MISSION
Research That Strengthens Human Learning
Learnology Labs conducts research at the intersection of the learning sciences, human development, and artificial intelligence. Our work is organized around four research pillars and a question that runs through all of them: can technology be designed to strengthen, rather than substitute for, the human relationships through which learning develops? Our findings to date suggest it can, with effects worth taking seriously.
Where Knowledge, Learning, and AI Connect
Across our portfolio, we study how knowledge builds, how it shapes learning, and how AI can be designed to strengthen the human relationships at the heart of development.
Our Four Research Pillars
AI & the Science of Personalization
Building research-grounded knowledge models that help AI identify what learners know, what they are ready to learn next, and how personalization can actually improve outcomes.
Ecosystems of Learning
Studying how children learn across connected environments — at home, in classrooms, with teachers, caregivers, materials, and the everyday moments that shape development.
Mathematical Learning Across Development
Mapping how mathematical knowledge builds from early childhood through algebra, so strong foundations can form before learning gaps widen.
Thinking, Learning, and Working in the Age of AI
Exploring how humans and AI can work together in ways that strengthen critical thinking, preserve productive struggle, and build human capacity.
PILLAR 01
AI & the Science of Personalization
Making AI-driven learning actually work.
Decades ago, Benjamin Bloom showed that one-to-one tutoring produces learning gains two standard deviations beyond conventional instruction, and challenged the field to achieve those results at scale. Despite years of educational technology, most digital interventions still fall far short. A growing body of research, including our own, points to why: personalization fails without the right infrastructure underneath it.
Effective personalization requires comprehensive, research-grounded knowledge models: granular maps of what there is to know in a domain, with the precursor and successor relationships among concepts explicitly encoded, so that a system can determine what a learner knows, doesn’t know, and is most ready to learn next. We develop the theory of these models, build them, and test them in rigorous field studies. Our findings to date suggest that systems built on this architecture can produce learning gains several times larger than typical edtech effects (Betts, Ryon, & Laski, 2026).
Our work also documents what happens without this infrastructure: our analysis of how generative AI systems make structurally predictable errors in early mathematics, crossing developmental boundaries, conflating concepts, and delivering misconceptions with confident polish (“The AI Mirror,” The Learning Agency, 2025).
KEY DEFINITION
Content-structural personalization
Making AI-driven learning actually work.
Decades ago, Benjamin Bloom showed that one-to-one tutoring produces learning gains two standard deviations beyond conventional instruction, and challenged the field to achieve those results at scale. Despite years of educational technology, most digital interventions still fall far short. A growing body of research, including our own, points to why: personalization fails without the right infrastructure underneath it.
Content-structural personalization is personalization driven by the structure of the knowledge itself. The approach is grounded in Knowledge Space Theory (Doignon & Falmagne, 1985), which provides the formal framework for representing a domain as a network of knowledge states governed by prerequisite relationships. Most adaptive systems personalize based on surface performance: what the learner just answered correctly or incorrectly. Content-structural personalization instead reasons over a formal map of the domain, with knowledge broken into fine-grained units and the precursor and successor relationships among them explicitly encoded. This allows a system to infer not just how a learner is performing, but which understandings they have, which they lack, and which they are most ready to build next
Selected Current Work
Early childhood mathematics knowledge model, ages 0–8 (operational; the research foundation for our ongoing field studies)
Early childhood literacy knowledge model, ages 0–8 (in development)
Knowledge infrastructure for algebra learning (proposal under review)
Theory of content-structural personalization (PME-NA 2026, accepted)
Documentation of generative AI failure modes in early mathematics (published, 2025)
PILLAR 02
Ecosystems of Learning
The child, the materials, the teacher, the home, and how learning accelerates when you reach all four.
Children don’t learn in apps; they learn in ecosystems: at the dinner table, in the classroom, in the conversations between home and school. In his classic work on the 2-sigma problem, Bloom identified the alterable objects of change in a child’s learning: the learner, the instructional materials, the teacher, and the home environment. Most educational improvement efforts work on only one or two, typically the teacher and the materials. Very few personalize to the needs of each individual child, and fewer still treat families as the powerful learning partners the evidence shows they can be. Our research asks what becomes possible when you work through all four at once.
Our current studies center on the most neglected objects: the child and the home. Foundational to this work is the RESET framework (Betts, 2021, 2024), which identifies five factors shaping parents’ engagement in their children’s learning (Role, Expectations, Skills, Efficacy, and Time) and the related construct of Math Parenting Identity. In our most recent field study, children’s math growth increased with the number of developmental messages their parents received, even though parents rarely completed the linked activities. The messages themselves appear to do work: changing what parents notice, say, and do in everyday moments. Informing adults may be as powerful a lever as prescribing activities for them.
Learning happens in ecosystems
The research centers on four connected objects of change: the child, the instructional materials, the teacher, and the home environment.
A growing branch of this work addresses the teacher. Early childhood and early elementary teachers typically receive minimal preparation in mathematics, and many enter classrooms without the content knowledge and pedagogical expertise needed to guide the math learning that sets children’s trajectories for later achievement. We are investigating how AI can address both halves of this problem: supporting teachers in the moment through an AI agent constrained by a research-grounded knowledge infrastructure, and building teachers’ own lasting knowledge, skill, and confidence over time. In one current study, teachers record informal observations of children in their classrooms. Because of the training gap, teachers don’t always know what they are seeing, what it means, or what to conclude from it. We are examining whether a knowledge-model-constrained agent can translate those raw observations into developmentally grounded insight: what each observation reveals about a child’s understanding, and what that child is most ready to learn next.
Much of our research in this pillar is conducted in studies of PAL (Personal Assistant for Learning), an adaptive learning system developed by Learnology.ai, an independent technology company. Learnology Labs leads efficacy research on PAL, conducted under external IRB oversight and using standardized independent outcome measures, in collaboration with the Thinking and Learning Lab at Boston College.
Selected Current Work
- Randomized pilot of parent-facing developmental nudges, pre-K (published: Betts, Ryon, & Laski, 2026)
- Dosage-response study, Junior-K through Grade 1 (manuscript in preparation; target AERA 2027)
- Multi-site field implementations involving 200+ children (underway)
- Co-design research with parents on nudge engagement and customization, across three institutionally distinct school contexts including Seven Arrows Elementary (NSF planning proposal, submitted)
- Teacher-facing observation and capacity-building tools (in development; classroom and family pilots beginning school year 2026–27)
PILLAR 03
Mathematical Learning Across Development
How mathematical knowledge builds, from first words about quantity through algebra, and why it matters far beyond the classroom.
Early mathematical knowledge is among the strongest predictors of later academic achievement, predicting not just later math performance but reading achievement, high school completion, and college attendance, even after controlling for family background. And the stakes extend beyond school: in a world where citizens are asked daily to interpret statistics, evaluate risk, and reason about data, innumeracy is not just an educational problem but a civic one. A society’s mathematical health is built, or lost, in its earliest years.
Mathematics is also the gateway to STEM. A child’s early mathematical trajectory shapes their access to science, technology, engineering, and mathematics pathways for decades afterward, which makes early math not only an educational and civic concern but a national workforce one. Our research in this pillar contributes to the science base for STEM readiness, beginning where STEM trajectories actually begin: before kindergarten.
Our research in this pillar contributes to the science base for STEM readiness, beginning where STEM trajectories actually begin: before kindergarten. Yet mathematics is structurally cumulative: later understanding depends on earlier foundations, and gaps compound silently for years before surfacing as “math difficulty.” Our research maps how mathematical knowledge actually develops, at a granularity fine enough to act on, and challenges the field’s persistent underestimation of early mathematics, a problem we’ve described as the triple expertise gap: practitioners undertrained in early math, a society that perceives it as simple, and too few specialists to build the knowledge infrastructure that educators and AI systems both need.
And the stakes extend beyond school: in a world where citizens are asked daily to interpret statistics, evaluate risk, and reason about data, innumeracy is not just an educational problem but a civic one. A society’s mathematical health is built, or lost, in its earliest years. Mathematics is also the gateway to STEM. A child’s early mathematical trajectory shapes their access to science, technology, engineering, and mathematics pathways for decades afterward, which makes early math not only an educational and civic concern but a national workforce one.
Mathematical Wellness
Mathematical wellness means building strong foundations before gaps form, rather than waiting to remediate after learning gaps have widened.
Selected Current Work
- Measurement of early mathematics growth using the TEMA-3, the gold-standard instrument for ages 3 through 8, as the pre/post outcome measure across all of our field studies (ongoing)
- Independent validation research on a multimodal, knowledge-state adaptive assessment of early mathematics, designed for pre-readers and benchmarked against the TEMA-3 (studies in design)
- Construction and empirical validation of fine-grained early mathematics learning trajectories (ongoing)
- Algebra learning progressions and misconception taxonomy (proposal under review)
- Public scholarship on early mathematics expertise gaps (published, 2025)
Our research in this pillar contributes to the science base for STEM readiness, beginning where STEM trajectories actually begin: before kindergarten. Yet mathematics is structurally cumulative: later understanding depends on earlier foundations, and gaps compound silently for years before surfacing as “math difficulty.” Our research maps how mathematical knowledge actually develops, at a granularity fine enough to act on, and challenges the field’s persistent underestimation of early mathematics, a problem we’ve described as the triple expertise gap: practitioners undertrained in early math, a society that perceives it as simple, and too few specialists to build the knowledge infrastructure that educators and AI systems both need.
- Measurement of early mathematics growth using the TEMA-3, the gold-standard instrument for ages 3 through 8, as the pre/post outcome measure across all of our field studies (ongoing)
- Independent validation research on a multimodal, knowledge-state adaptive assessment of early mathematics, designed for pre-readers and benchmarked against the TEMA-3 (studies in design)
- Construction and empirical validation of fine-grained early mathematics learning trajectories (ongoing)
- Algebra learning progressions and misconception taxonomy (proposal under review)
- Public scholarship on early mathematics expertise gaps (published, 2025)
Yet mathematics is structurally cumulative: later understanding depends on earlier foundations, and gaps compound silently for years before surfacing as “math difficulty.” Our research maps how mathematical knowledge actually develops, at a granularity fine enough to act on, and challenges the field’s persistent underestimation of early mathematics, a problem we’ve described as the triple expertise gap: practitioners undertrained in early math, a society that perceives it as simple, and too few specialists to build the knowledge infrastructure that educators and AI systems both need. Across this pillar runs a prevention orientation we call mathematical wellness: building strong foundations before gaps form, rather than remediating after they widen.
Selected Current Work
Measurement of early mathematics growth using the TEMA-3, the gold-standard instrument for ages 3 through 8, as the pre/post outcome measure across all of our field studies (ongoing)
Independent validation research on a multimodal, knowledge-state adaptive assessment of early mathematics, designed for pre-readers and benchmarked against the TEMA-3 (studies in design)
Construction and empirical validation of fine-grained early mathematics learning trajectories (ongoing)
Algebra learning progressions and misconception taxonomy (proposal under review)
Public scholarship on early mathematics expertise gaps (published, 2025)
PILLAR 03
Thinking, Learning, and Working in the Age of AI
What uniquely human expertise must we cultivate, and what should we delegate to machines?
As AI grows capable of producing fluent answers to nearly any question, the most urgent questions in education shift beneath our feet. What are thinking and learning, really? How do we preserve productive struggle, the effortful engagement through which deep understanding develops, in a world of instant answers? And what capacities will the next generation need to live, learn, and work well in a world increasingly shaped by AI?
Our research in this pillar starts from the premise that effective human-AI partnership is itself a form of expertise, one that must be deliberately developed, not assumed. Drawing on the theory of distributed cognition, we study how the cognitive work of learning and teaching can be productively divided: humans contributing what humans do best (relational, responsive, judgment-rich engagement with other minds) while machines manage complexity beyond the reach of any single human mind. Within this framework we examine both sides of the partnership: how AI systems can be designed to build human capacity rather than erode it, and what knowledge, dispositions, and self-regulation humans need to be effective partners, and effective thinkers, in AI-saturated environments.
This pillar also includes our contribution to a multi-organization research working group, part of the Human Intelligences initiative, developing a layered model of critical thinking for the AI era: one that treats it not as a checklist of analytic skills but as an emergent capacity built from cognitive subskills, epistemic competencies, and intrapersonal foundations.
Selected Current Work
AI that builds human capacity: studies of AI-supported development of teacher and caregiver expertise (in development; pilots beginning school year 2026–27)
Layered model of critical thinking for the AI era (multi-organization working group; in progress)
Collaborative intelligence and distributed cognition in education (published, 2025)
Optimizing for learning growth vs. mastery verification: ZPD elasticity (published, 2024)
How We Work
Research at Learnology Labs happens through partnership. Our studies are designed with university researchers, conducted in real classrooms and homes, built alongside technologists, and carried into policy conversations. These relationships are not add-ons to our research; they are how rigorous, practically relevant science gets done.
Boston College:
Thinking and Learning Lab. Our closest academic partnership is with the Thinking and Learning Lab at Boston College, directed by Dr. Elida V. Laski, who serves as Senior Research Advisor to Learnology Labs. Dr. Laski is embedded in our research from the start, shaping study design, knowledge modeling, and analysis rather than evaluating finished work from the outside. Our field studies are conducted under IRB approval through Boston College, and Dr. Laski is a co-author on our published research. [PENDING DR. LASKI’S WRITTEN APPROVAL OF THIS PARAGRAPH]
Early Learning Coalition of Palm Beach County:
Our community research partnership with ELCPBC, which serves primarily low-income families in one of the most linguistically diverse counties in the United States, grounds our work in the populations that stand to benefit most. ELCPBC was the site of our first randomized pilot study, and our collaboration continues to grow.
School Partners:
School partners. Our field research is conducted with school partners across three states, including Seven Arrows Elementary School in California and the Interboro School District in Pennsylvania. These partnerships let us study learning where it actually happens: in classrooms and family routines, across institutionally and demographically distinct communities.
Learnology.ai:
While Learnology Labs is an independent 501(c)(3) nonprofit research institute, we partner closely with Learnology.ai, a learning-science-first technology company built on a simple premise: that educational technology should begin with the science of learning, not retrofit it after the fact. Where most edtech starts with a product and looks for evidence later, Learnology.ai starts with the research, building its smart learning systems, including PAL, from research-grounded knowledge models, learning theory, and evidence-centered design from the ground up. Labs’ research directly informs that work, and in turn, Labs conducts efficacy research on the resulting systems under external IRB oversight, using standardized independent outcome measures, in collaboration with the Thinking and Learning Lab at Boston College. This research–practice–developer cycle, which we have described in our published work, is our model for how evidence and innovation should move together.
Note:
We also collaborate with researchers and organizations across the field, including Mindset Copilot and the Human Intelligences initiative, and we welcome new research partnerships. To explore working together, contact [email protected].
RESEARCH OUTPUT
Selected Bibliography
Learnology Labs Research Outputs: Peer-Reviewed & Conference Papers
Betts, A., Ryon, B., Laski, E. V., & Gunderia, S. (in press). Designing for the just-right moment: Driving human behavior through temporal scaffolding in PAL. In Proceedings of HCI International 2026. Springer.
Betts, A., Gunderia, S., & Pullen, P. (in press). The evidentiary chain: Extending Mislevy’s evidence-centered design across learning for children, educators, and families. In E. Tucker & M. E. Oliveri (Eds.), Modeling What Matters: The Research and Legacy of Robert J. Mislevy. [add publisher at proofs]
Betts, A., & Hughes, D. (accepted). The promise of personalization: Operationalizing mathematical knowledge models in the age of AI. PME-NA 48, 2026.
Betts, A., Ryon, B., Laski, E. V., & Gunderia, S. (2026). Agentic PAL: Designing human-empowered AI partnerships for early childhood mathematics learning. In Proceedings of the Learning Engineering Research Network Convening (LERN 2026). EdTech Archives. https://doi.org/10.59668/2551.25418
Betts, A., Ryon, B., & Laski, E. V. (2026). “Smart” nudges, real gains: Family engagement for kindergarten math readiness. Paper presented at the Annual Meeting of the American Educational Research Association, Los Angeles, CA.
Betts, A., Ryon, B., Laski, E. V., & Gunderia, S. (2026). Learning in the wild: PAL’s new tools and methods for research in early childhood mathematics. International Society of the Learning Sciences.
Betts, A., Ryon, B., Laski, E. V., Gunderia, S., & Hughes, D. (2026). Partnering with purpose: A research–practice–developer model for evidence-driven innovation. International Society of the Learning Sciences.
Betts, A., Laski, E. V., & Ryon, B. (2026). When the child is not the user: Leveraging AI to support early childhood well-being via adult–child interactions. Interaction Design and Children (IDC).
Betts, A., Ryon, B., & Laski, E. V. (2025). Beyond screens: Human-mediated “smart” systems for early math learning. PME-NA 47, State College, PA.
Betts, A., Gunderia, S., Hughes, D., Owen, L., & Bang, H. J. (2025). Beyond measurement: Assessment as a catalyst for personalizing learning and improving outcomes. In Handbook on Assessment in the Service of Learning, Vol. III (pp. 383–415). Springer.
Betts, A., Hughes, D., & Gunderia, S. (2024). The best start: A Bloomsian perspective on AI-powered innovations in mathematics education. PME-NA 46.
In Preparation
Betts, A., Ryon, B., & Laski, E. V. (manuscript in preparation). Every message matters: PAL dosage exposure and children’s early mathematics growth. Target: AERA 2027.
Learnology Labs Policy & Public Scholarship
Betts, A., Gunderia, S., Hughes, D., & Lenihan, E. (2024). GenAI in Education Research Accelerator (GenAiRA). Federation of American Scientists, Day One Project.
Betts, A. (2025). The AI mirror: How GenAI reflects and amplifies gaps in early math expertise. The Cutting Ed, The Learning Agency.
Betts, A. (2025). Human-AI partnerships in education: Entering the age of collaborative intelligence. The Cutting Ed, The Learning Agency.
Betts, A. (2025). Learning vs. mastery: Rethinking “smart” learning systems for optimal growth. LinkedIn.
Betts, A. (2024). Reimagining education: How knowledge models and AI can help teachers address the learner variability challenge. Learnology Labs.
Betts, A. (2024). The path to AI-driven learning: Building critical knowledge infrastructure. Learnology Labs.
Betts, A. (2024). The knowledge model imperative: Why human expertise is essential for AI in education. Learnology Labs.
Betts, A. (2024). The future is now: Accelerating learning through knowledge space theory and AI-driven personalization. Learnology Labs.
Foundational Work by Our Researchers (prior to or outside Learnology Labs; selected)
The research program at Learnology Labs builds on a foundation of earlier scholarship by its researchers.
Betts, A. (2024). Examining critical factors in parent–child math engagement. Doctoral dissertation, State University of New York at Buffalo.
Betts, A., Son, J.-W., & Bang, H.-J. (2024). Dismantling deficit-based perspectives of the pre-primary home math environments of African American and Multiracial families. AERA Annual Meeting.
Betts, A., Hughes, D., Plache, L., & Smith, K. (2024). Stretching the zone of proximal development: Accelerating learning through ZPD elasticity. IAFOR International Conference on Education.
Betts, A., & Son, J.-W. (2024). Toward an understanding of “Math Parenting Identity”: Parent perceptions of the home math environments of young children. International Society of the Learning Sciences.
Betts, A., Son, J.-W., & Bang, H.-J. (2023). Learning to parent mathematically: Critical factors in parent–child math engagement. PME-NA 45.
Betts, A., & Son, J.-W. (2022). Why parents do what they do: Developing and validating a survey for the mathematical lives of parents and children. Paris Conference on Education.
Betts, A., & Thai, K. P. (Eds.). (2022). Handbook of Research on Innovative Approaches to Early Childhood Education and School Readiness. IGI Publishing.
Thai, K. P., Betts, A., & Gunderia, S. (2022). Personalized mastery-based learning ecosystem: A new paradigm for improving outcomes and defying expectations in early childhood. In Handbook of Research on Innovative Approaches to Early Childhood Education and School Readiness (pp. 665–694). IGI Publishing.
Betts, A., Thai, K. P., & Gunderia, S. (2021). Personalized Mastery Learning Ecosystems (PMLE): Using Bloom’s four objects of change to drive learning in Adaptive Instructional Systems. In HCII 2021 Proceedings, LNCS (pp. 29–52). Springer.
Betts, A. (2021). The RESET framework: Examining critical factors in parent–child math participation. IAFOR International Conference on Education.
Betts, A., Thai, K. P., Gunderia, S., Hidalgo, P., Rothschild, M., & Hughes, D. (2020). An ambient and pervasive personalized learning ecosystem (APPLE): “Smart learning” in the age of the Internet of Things. HCII 2020 Proceedings, LNCS. Springer.
Betts, A., & Son, J.-W. (2020). Fostering parent–child math talk with the 4Cs. Mathematics Teacher: Learning and Teaching PK–12. NCTM.
Betts, A. (2019). Mastery learning in early childhood mathematics through adaptive technologies. IAFOR International Conference on Education.
Patents
Dohring, D. C., Hendry, D. A., Gunderia, S., Hughes, D., Owen, V. E., Jacobs, D. E., Betts, A., & Salak, W. (2022). Personalized mastery learning platforms, systems, media, and methods (U.S. Patent No. 11,380,211). U.S. Patent and Trademark Office.
Dohring, D. C., Hendry, D. A., Gunderia, S., Hughes, D., Owen, V. E., Jacobs, D. E., Betts, A., & Salak, W. (2021). System and method for dynamically editing online interactive elements architecture (U.S. Patent No. 11,151,887). U.S. Patent and Trademark Office.
Dohring, D. C., Hendry, D. A., Gunderia, S., Hughes, D., Owen, V. E., Jacobs, D. E., Betts, A., & Salak, W. (2019). Personalized mastery learning platforms, systems, media, and methods (U.S. Patent No. 10,490,092). U.S. Patent and Trademark Office.
PARTNER WITH US
Drive Change Through Research
Join leading institutions in shaping the future of education. We provide the rigorous evidence base required to scale impactful learning models.
Together, we translate academic rigor into scalable impact.
An independent nonprofit research institute dedicated to understanding how people learn and translating that research into evidence-based tools, programs, and policy.
Institutional Details
Learnology Labs is an independent 501(c)(3) nonprofit research institute.
Contributions are tax-deductible to the extent allowed by law.
© Copyright Learnology labs 2026. All rights reserved.