Synthetic Users in More Industries
Synthetic Users in U.S. Higher Education
By Sherry Jones (September 2025)
Synthetic users represent a significant advancement in artificial intelligence, offering sophisticated simulations of human behavior that are transforming education. These AI-generated personas replicate thoughts, behaviors, and responses of real participants, moving far beyond basic scripted bots.
What Are Synthetic Users?
Advanced AI Personas
Unlike rigid, rule-based programs, synthetic users leverage advanced natural language processing, behavioral modeling, and contextual awareness to engage in fluid, open-ended dialogue.
They adjust responses based on interaction context and mimic realistic human emotions, motivations, and complex decision-making patterns.
These software agents are trained on extensive datasets, including historical user journey data, eye-tracking archives, interaction heatmaps, and cognitive load metrics, to closely approximate authentic human usage patterns.
Synthetic Data vs. Synthetic Users
Synthetic Data
Artificially generated datasets designed to imitate statistical properties of real-world data while limiting information about individual records. Primary purpose is privacy preservation while allowing robust analysis.
Synthetic Users
AI-powered agents that interact or simulate behaviors. Can be trained on synthetic data to enhance realism or generate synthetic data through simulated interactions.
Both fall under privacy-enhancing technologies (PETs) that mitigate disclosure risks associated with sensitive information. While synthetic data provides the anonymized information base, synthetic users provide the dynamic, interactive simulation layer.
Key Benefits in Education
Speed & Cost Efficiency
Replicate real user interactions for extensive testing in a fraction of the time required with human participants, bypassing recruitment bottlenecks.
Privacy Preservation
Eliminates concerns surrounding consent, data ownership, and privacy protection - invaluable for FERPA compliance in educational contexts.
Consistency & Reproducibility
Provides neutral responses unaffected by social cues or researcher presence, reducing variables in experimental designs.
Scalability & Diversity
Enables diverse and inclusive participant sets, including segments that would be costly or time-consuming to recruit manually.
Learning Analytics and Privacy Protection
U.S. colleges and universities increasingly use synthetic data to conduct learning analytics while protecting student privacy. The Unizin Consortium created a synthetic data warehouse that mirrors real Canvas LMS data, generating fake students, courses, and clickstream events with the same statistical patterns as real university data.
MIT researchers built a Synthetic Data Vault (SDV) to generate realistic student-course datasets. The SDV creates "synthetic students" with fake academic records that statistically resemble real data, enabling predictive models without using protected student records. The SDV has been downloaded over a million times.
MIT's Synthetic Data Success
1M+
Downloads
SDV downloaded over a million times by universities and companies
0%
Accuracy Loss
No significant difference in predictive model accuracy compared to real data
100%
Privacy Protected
Complete elimination of identifiable student information
Educational Research Applications
Computing Education
Leinonen et al. (2024) used large language models to generate synthetic code submissions for programming exercises. The LLM-generated student programs, including common bugs, matched real student error distributions closely.
Survey Research
The SQRA framework creates synthetic student personas that answer draft survey questions using LLMs. Researchers analyze AI-AI conversations to stress-test questions for clarity and bias before involving real participants.
AI-Powered Tutoring Systems
Synthetic users are instrumental in developing personalized learning platforms. By simulating diverse learning behaviors, synthetic data serves as training ground for AI models to identify learning gaps and tailor interventions effectively.
Praxis AI partners with universities to deploy AI "digital twins" of professors using Anthropic's LLMs. Each instructor completes a personality quiz, and the system creates an AI agent that knows course material and the instructor's style. Students can ask questions anytime.
Synthesis Tutor: Personalized Math Learning
Synthesis Tutor exemplifies personalized learning companions that dynamically adapt to a child's ability. Unlike traditional rote memorization tools, it focuses on building deep foundational understanding through multi-sensory, engaging, and gamified experiences.
Initially accessible to SpaceX families, it now serves a broader community covering elementary math for ages 5-11. The system reads content aloud for emerging readers and is particularly effective with neurodiverse students.
Synthesis Tutor in Action
This video demonstrates how Synthesis Tutor moves beyond traditional methods, offering a truly personalized journey for each child. It highlights the AI's ability to instantly assess understanding and adapt the curriculum, ensuring optimal challenge and support.
Dynamic Adaptation
The AI continuously analyzes a child's responses and learning patterns, adjusting the pace and complexity of problems in real-time. This ensures that content is always at the perfect level, preventing boredom or frustration.
Foundational Mastery
Unlike rote memorization, Synthesis Tutor emphasizes building a deep conceptual understanding. It uses visual, auditory, and interactive elements to explain concepts thoroughly before progressing.
Engaging & Gamified
Learning is transformed into an adventure through gamified challenges, rewards, and multi-sensory feedback. This approach keeps children motivated and excited to learn, turning math into an enjoyable activity.
Inclusive Accessibility
With features like text-to-speech for emerging readers and carefully designed interfaces, the platform is effective for a diverse range of learners, including neurodiverse students, fostering confidence and capability in math.
Amira Learning: AI Reading Coach
Real-Time Assessment
Advanced speech recognition continuously assesses reading mastery as students read aloud, identifying difficulties instantly.
Micro-Interventions
Provides immediate interventions based on decades of reading and neuroscience research.
Proven Results
Minimum seven additional weeks of reading growth with less than 30 minutes weekly use.
Amira Learning AI Reading Coach in Action
Watch this video to see Amira in action: Amira Learning Demo. The video demonstrates how Amira's AI listens to students read aloud, provides real-time feedback and corrective prompts, and tracks their progress to pinpoint specific areas of improvement in phonics, fluency, and comprehension.
Advanced Speech AI & Diagnostics
Amira's sophisticated speech recognition engine accurately tracks hundreds of reading behaviors per minute, identifying specific struggles with decoding, phonics, and prosody. This allows for precise, diagnostic assessment far beyond traditional methods.
Personalized Micro-Interventions
When a student falters, Amira delivers immediate, targeted support. Whether it's sounding out a word, rereading a sentence for fluency, or asking a comprehension question, interventions are tailored to the individual's needs, preventing the embedding of incorrect habits.
Data-Driven Insights for Educators
Amira compiles detailed reports for teachers, highlighting student growth, areas of difficulty, and recommended next steps. This data empowers educators to make informed instructional decisions and provide targeted support in the classroom.
Engaging & Adaptive Learning Journey
Designed to keep students motivated, Amira adapts to each child's pace and skill level, offering a supportive and non-judgmental environment. This personalized approach fosters confidence and makes reading practice enjoyable.
Praxis AI: Digital Professor Twins
75%
Students interacted with professor's AI twin
14%
Interaction rate with generic chatbots
1
Full letter grade improvement (C to B)
Early results at Clemson and Alabama State show dramatic engagement gains. This "round-the-clock tutor" approach improved learning outcomes while freeing professors from routine queries to focus on high-value teaching.
Praxis AI Digital Professor Twin in Action
As demonstrated in the video, modern AI tutoring systems move beyond simple Q&A to provide interactive explanations, personalized feedback, and adaptive learning paths. Praxis AI embodies this evolution by creating a seamless extension of the professor, ensuring students receive timely, accurate, and empathetic support tailored to their individual needs, ultimately fostering a more engaging and effective educational environment.
Digital Twin Students: The Future
James Cook University in Australia is developing an AI "twin" for each student that aggregates LMS, enrollment, and background data to provide personalized recommendations and nudges.
Like a fitness tracker for learning, these digital twins promise a future where LLM-driven agents proactively support each student's learning journey, dramatically improving retention and engagement.
Platform Development and Quality Assurance
Edtech vendors and universities employ synthetic users to test and improve digital learning platforms. By generating large populations of fake learners with varied profiles, product teams can stress-test features without involving real students.
Industry practices from tech companies like Google and Meta are migrating into higher education. Developers use AI "bot students" to simulate thousands of interactions on new quiz systems or discussion forums before rollout, enabling rapid iteration while preserving student privacy.
Medical Education Simulations
MedVR Education's AI-Humans
Creates immersive, AI-based patient assessment scenarios for medical professionals. Features advanced AI avatar technology with dynamic facial expressions, real-time emotion recognition, and human-like voices that adapt to convey medical situation gravity.
Real-Time AI-Driven Assessment analyzes conversations against pre-set rubrics, providing immediate debriefing, personalized feedback, and scores for improved patient care outcomes.
MedVR Medical Education Simulations in Action
This video illustrates the dynamic capabilities of AI in simulating complex patient interactions. It showcases how virtual patients can exhibit realistic emotional responses and adapt their behavior based on the learner's actions, providing a truly immersive and challenging experience for assessing clinical judgment and interpersonal skills.
Dynamic AI Patient Avatars
These avatars move beyond static simulations, featuring realistic facial expressions, real-time emotion recognition, and human-like voices that adjust to reflect the gravity of medical situations. This fosters empathetic communication and accurate assessment.
Real-Time Performance Analytics
AI systems continuously analyze conversations and clinical decisions against pre-set rubrics. This immediate assessment provides detailed debriefing, personalized feedback, and objective scores, leading to continuous improvement in patient care outcomes.
Adaptive Learning Pathways
AI enables scenarios to dynamically adjust difficulty and complexity based on the learner's performance. This ensures each trainee receives a customized learning experience, targeting specific areas for development and accelerating skill acquisition.
Business Education Simulations
01
AI Integration
Queen Mary University integrates AI into business simulations through structured four-part scaffolding approach using Microsoft Copilot and ChatGPT.
02
Strategic Analysis
Students generate mission statements, SWOT analyses, and analyze complex simulation data for strategic solutions.
03
Skill Development
Enhances AI literacy, strengthens critical thinking, and accelerates decision-making skills essential for evolving job market.
Entrepreneurship Training with AI
ExEC AI Simulators
AI Interviewing Simulator allows students to "interview" an AI acting as potential customer, receiving realistic responses and personalized feedback on questioning strategies. This reduces student anxiety and boosts engagement.
Financial Projection Simulator evaluates business assumptions like price, costs, and sales channels, providing immediate feedback on how decisions impact revenue and profit.
ExEC AI Interviewing Simulator: See it in Action
Watch a full demonstration of the simulator here: ExEC AI Interviewing Simulator Video
How the Simulator Works
Realistic AI Customer Responses
Students engage with an AI acting as a potential customer, receiving dynamic and realistic responses that mirror genuine human interaction. This allows for an authentic practice experience.
Personalized Feedback & Analytics
The simulator provides immediate, detailed feedback on questioning strategies, communication clarity, active listening, and problem identification. This helps students refine their approach quickly.
Reduced Anxiety & Enhanced Engagement
Practicing in a low-stakes, confidential environment helps students overcome the anxiety associated with initial customer interviews, boosting their confidence and overall engagement in the learning process.
Iterative Skill Development
Students can repeatedly practice and iterate on their interview techniques, experimenting with different questions and approaches to find what resonates best with potential customers.
Aquifer: Virtual Patient Cases
15M+
Cases Completed
Virtual patient cases completed since 2006
2006
Since
Years of providing clinical learning resources
Aquifer provides extensive online clinical learning resources for medical and health professions. These deep, realistic cases are designed to develop clinical reasoning skills and include customizable clinical quizzes for comprehensive medical education.
Future Directions: Generative AI Integration
Simulated Classrooms
AI will simulate entire classroom discussions and peer review interactions, creating synthetic student cohorts for scenario planning.
Complete Digital Twins
Holistic virtual replicas of learning environments with AI tutors that understand knowledge gaps, motivations, and context.
Hyper-Personalization
AI companions will provide hyper-personalized learning at scale, transforming how students learn and educators design courses.
Policy and Governance Considerations
Emerging Challenges
New roles are emerging as faculty and support staff curate and train synthetic user agents, requiring oversight of data and ethics. Institutions need policies alongside FERPA compliance to govern synthetic-user tools.
Ensuring AI-driven tutors are transparent about their nature and that student interactions are secured becomes critical for widespread adoption.

Experts warn that while AI companions can powerfully personalize education, they raise concerns about data governance and integrity that institutions must address proactively.
The Transformation of Higher Education
The trajectory is clear: generative AI and synthetic user modeling will become mainstream in higher education. These technologies enable hyper-personalized learning at scale while transforming both how students learn and how educators design courses.
Synthetic users represent not just technological convenience but a necessary component for advancing data-driven initiatives in education responsibly. As institutions prioritize data security and ethical considerations, they position themselves at the forefront of educational innovation.
Support, Not Replace
AI serves as augmentative tool, enhancing human capabilities rather than replacing educators
Privacy-First Innovation
Enables data-driven research while respecting student privacy and regulatory compliance
Scalable Personalization
Provides individualized learning experiences at unprecedented scale and efficiency
References
[1] (2024 September 10). Leveraging the Unizin Data Platform to Support Student Learning. Unizin. https://unizin.org/2024/09/11/leveraging-the-unizin-data-platform-to-support-student-learning/
[3] [4] Eastwood, B. (2023 January 23). What is synthetic data — and how can it help you competitively? MIT Sloan. https://mitsloan.mit.edu/ideas-made-to-matter/what-synthetic-data-and-how-can-it-help-you-competitively
[5] Leinonen, J., Denny, P., Kilijunen, O., MacNeil, S., Sarsa, S., & Hellas, A. (2024 November 1). LLM-itation is the Sincerest Form of Data: Generating Synthetic Buggy Code Submissions for Computing Education. arXiv. https://arxiv.org/abs/2411.10455
[6] Mburu, T.K., Rong, K., McColley, C.J., Werth, A., Rong, C.J., Werth, K., & McColley, A. (2025 May 2). Methodological Foundations for AI-Driven Survey Question Generation. ArXiv. https://arxiv.org/html/2505.01150v1
[7] Rosala, M. & Moran, K. (2024 June 21). Synthetic Users: If, When, and How to Use AI-Generated “Research.” NN/G. https://www.nngroup.com/articles/synthetic-users/
[8] [9] [10] Wood, C. (2025, March 6). Professors' AI twins loosen schedules, boost grades. EdScoop. https://edscoop.com/digital-twin-ai-university-praxis-anthropic/
[11] [12] de Loryn, B. (2025). Digital Twin Student: Personalising the student learning experience. JCU Australia. https://www.jcu.edu.au/jcu-ideas-lab/blog/digital-twin-student
[13] Mart, A. (2025, July 21). What Are Synthetic Users and How Do They Work? M1-Project. https://www.m1-project.com/blog/what-are-synthetic-users-and-how-do-they-work
[14] Wood, C. (2025, March 6). Professors' AI twins loosen schedules, boost grades. EdScoop. https://edscoop.com/digital-twin-ai-university-praxis-anthropic/
[15] Rosala, M. & Moran, K. (2024 June 21). Synthetic Users: If, When, and How to Use AI-Generated “Research.” NN/G. https://www.nngroup.com/articles/synthetic-users/
[17] Arellano, B. (2025). Using AI for Curriculum Design + 21 AI Prompt Suggestions. Fly Plugins. 12 https://flyplugins.com/ai-for-curriculum-design/
[18] Gurung, T.W. (2025 August 15). What Is the Impact of Using Synthetic Users on Research? Qualz.ai. 1 https://qualz.ai/blogs/impact-of-using-synthetic-users/
[19] Smith, A. (2025, June 24). Synthetic Users: The Future of UX Research? Bootcamp. Medium. 3 https://medium.com/design-bootcamp/synthetic-users-the-future-of-ux-research-4111926fe3fa
[20] Seeman, J., Williams, A.R., & Bowen, C.M. (2025, January). Synthetic Data for the Nebraska Statewide Workforce & Educational Reporting System. Urban Institute. 8 https://www.urban.org/sites/default/files/2025-01/Synthetic-Data-for-the-Nebraska-Statewide-Workforce-%26-Educational-Reporting-System.pdf
[21] (2024). Synthesis Tutor. Synthesis. 13 https://www.synthesis.com/tutor
[22] (2025). Meet Amira. Amira Learning. 14 https://amiralearning.com/
[23] (2024). AI Humans Authoring Platform. MedVR Education. 16 https://medvr.education/ai-humans-ai-based-patient-assessment/
[24] (2020, July 1). Aquifer Practice Smart: Student Self-Assessment Question Bank & Quiz Generator. Aquifer. 17 https://aquifer.org/
[25] Woon, A. (2025, February 18). AI-Driven Simulations Build Decision-Making Skills. AACSB. 18 https://www.aacsb.edu/insights/articles/2025/02/ai-driven-simulations-build-decision-making-skills
[26] (2025, January 29). AI Simulators that Reduce Student Anxiety and Boost Engagement. Teaching Entrepreneurship. 19 https://www.teachingentrepreneurship.org/ai-simulators/