Thinkster Learning Prediqt.ai

Our patented AI solutions (AI Patent #1, AI Patent #2) leverages AI to transform assessment practices, fostering a more equitable, engaging, and data-driven approach to learning. Here's how our core components work together:

1. AI-Powered Question Tagging: 

Tag Types:  Questions are meticulously tagged using a multi-layered approach. 

Difficulty Level: Categorized from beginner to advanced, ensuring students receive appropriately challenging questions.

Comprehensibility: Analyzed for clarity and ease of understanding, considering factors like vocabulary and sentence structure.

Culturally and regionally Unbiased: Questions avoid references or scenarios specific to certain cultures, promoting inclusivity. Additionally, cultural tags can identify questions potentially biased against a particular student group.

Automated with Expert Oversight:  The tagging process is primarily automated for efficiency. Subject matter experts regularly review tag accuracy and address potential biases to ensure the system remains fair and effective.

2. Unbiased Assessment & Student Personas: 

Unbiased Question Pool: Students begin their learning journey with assessments from a curated pool of questions designed to be clear, straightforward, and free from cultural or regional bias.

Data Collection with Safeguards: Anonymized background data, including ethnicity, region, and language is collected with robust security measures in place. This data is essential for building student profiles but is never used in a way that could disadvantage any learner.

Correlation Analysis with Expert Review: Performance data is analyzed to identify potential correlations between student background and learning needs. A range of acceptable ‘P-Value’ is set in order to avoid stereotyping. A critical step involves subject-matter experts reviewing these correlations.

3. AI-powered Active Replay Technology (ART™): ART analyzes student work. It replays the solution steps that the student writes on the app while attempting a question. It can be utilized to catch errors/hesitation/lack of confidence in solving specific steps of a problem.

Roadmap: Optical Character Recognition (OCR): Using OCR we can scan each step of the solution and assign it a confidence score. For special needs students, answers can be submitted as verbal responses which can be converted into transcripts and analyzed by AI.

4. Personalized Learning Paths: Our patented AI engine analyzes student persona, question tags, and ongoing performance data to generate individualized learning paths.

Mastery-Based Learning & Adaptability: The system prioritizes mastery, ensuring students solidify understanding before progressing. Difficulty levels are adjusted dynamically, and targeted support is offered based on performance. 

5. Unbiased and Contextualized content: 

Selecting the perfect questions: We use a massive pool of assessment items to find the most relevant and unbiased ones that match the current learning level.

Making it relatable: AI can adjust the context of a question to make it easier for a student to understand. This reduces cognitive load and keeps them engaged.

6. Teacher Insights and Support: 

Detailed Performance Reports: Teachers receive comprehensive reports with student performance data at individual and concept/topic levels. This allows them to pinpoint areas where students excel or require additional support.

AI-Suggested Interventions & Teacher Override: AI suggests targeted remediation strategies for each student. However, teachers can customize interventions for individual students.

Our AI-powered solution addresses the specific needs of priority learners, including Black and Latino students, and students experiencing poverty. 

1. Traditional assessments can be culturally insensitive and disadvantage certain demographics. AI analysis helps create unbiased assessments and culturally-sensitive questions, leveling the playing field for priority learners.

Fair Question Pool: This ensures priority learners, who may have different backgrounds or experiences than the majority, aren't judged unfairly. They can showcase their understanding without being disadvantaged by culturally biased questions.

Making it relatable: AI can adjust the context of a question to make it easier for a priority learner to understand. This reduces any unnecessary cognitive load and keeps them engaged.

2.Our system analyzes student performance and adjusts instruction pace dynamically. Students who need extra time or targeted support receive it, ensuring a strong foundation. This builds confidence in priority learners who might often face negative stereotypes.

AI-powered Active Replay Technology (ART™):

Understanding Student Steps: For priority learners who might be hesitant to participate in class or struggle to explain their thought process, ART provides valuable insights. It can identify areas where they get stuck without needing them to speak up, allowing teachers to offer help.

Pinpointing Errors & Confidence Levels: Some students might make different types of errors than others, and ART can pinpoint these differences. With this knowledge, teachers can provide targeted interventions that address the specific learning gaps.

Personalized Learning Paths: ART personalizes the learning experience by focusing on the specific areas they struggle with. This targeted approach allows them to catch up and progress at their own pace.

Accessibility for All Learners: This is especially helpful for students who may have learning disabilities or language barriers. Being able to answer verbally removes a potential hurdle and allows them to show their understanding in a way that works for them.

3.Traditional testing formats can be intimidating. They highlight what students got wrong rather than their strengths and progress. This negative focus can be demotivating for priority learners who are already working hard to overcome challenges.

Engaging Assessments: Our platform incorporates playful elements, adaptive difficulty levels, and multiple attempts to make assessments enjoyable and reduce test anxiety.

Gamification: The system leverages gamification techniques, where students earn points and badges for progress, fostering a positive and motivating learning environment.

4.Large classes make it tough to tailor instruction for each student. Hidden learning gaps and communication barriers can further challenge identifying their needs. This creates an equity gap, requiring extra support to close it despite pressure for overall achievement.

Increased Efficiency: AI identifies patterns that might be missed by educators. It takes over time-consuming assessment analysis, freeing up valuable teacher time. This allows educators to focus on personalized instruction, relationship building, and addressing the specific needs.

Collaboration, not Automation: AI acts as a collaborative tool. It provides data-backed suggestions for remediation, concept reinforcement, and learning path adjustments. However, teachers retain decision-making authority, fostering a partnership between educators and technology.

Proven Expertise in EdTech (14+ Years):

• Our team has over 14 years of experience in the educational technology space. This extensive experience translates to a deep understanding of the educational landscape and the challenges faced by students and educators.

Generated Real Results for Parents/Consumers:

• We have hundreds of case studies and testimonials from parents of students whose learning experiences have been deeply transformed by their use of Thinkster.  This has translated into unheard levels of stickiness from parents who stay with us on average for 30+ months, a lifetime typically unheard of in the B2C tutoring market.  When parents see that the results we deliver to their students are transformative, transparent, sustainable, and auditable, they have no hesitation in continuing with our program.  Because of this, parents have also repeatedly requested that we add more subjects that we can provide tutoring in using our patented and proprietary technology stack.

Experienced Educators on Staff:  

• Our team comprises highly respected educators with extensive experience in curriculum development, content creation, and working effectively with diverse learners, including priority learners. This ensures our solutions are grounded in best practices and cater to the unique needs of all students.

• Understanding the Needs of Priority Learners: We prioritize understanding the specific needs of underserved communities, including Black and Latino learners and students experiencing poverty. We recognize the importance of equity in education and are committed to developing solutions that dismantle barriers and empower all students to succeed.

Revolutionizing Instruction with AI-powered (ART™): 

• We've developed our proprietary Active Replay Technology (ART) in-house. ART analyzes student work steps, pinpoints errors, and assigns confidence scores. Our near term plans include the ability to add OCR and AI analytics and heuristics to drive insights based on this ART.  This empowers educators with deep insights into student thinking, allowing them to personalize instruction, address specific needs, and provide targeted interventions for all learners, including priority learners.

Patented AI for Personalized Learning:

• We own multiple patents in AI specifically designed to curate personalized learning journeys for students. This technology ensures students are challenged yet engaged by optimally difficult assignments, fostering a growth mindset and maximizing learning potential.

Importance of Community Collaboration:

We firmly believe in the value of community input for shaping effective educational solutions. We are actively exploring avenues to integrate the perspectives and needs of the communities we serve into the design and implementation of this AI-powered assessment system.

We have served thousands of students in over 30 countries using our technology that includes two AI patents, have generated over $21M in revenues at 70% Gross Margins, with 4.85/5 star customer reviews, & have 4M+ visitors exploring our website MoM. Our target is the global tutoring market with $300B in annual spend.

We have raised over $12M, including Series A funding, from multiple marquee investors.

We have rave reviews from The Wall Street Journal, The New York Times, & others, strong customer average retention rate of 2.5+ years, & marketing investments boasting an ROI exceeding 300% in this timeframe.

Yes. Since we are available digitally and online, our solution is actively being used by tutors and students across all the states in the U.S. today.  We have also been approved recently as a vendor to sell solutions to schools in the State of Texas.

Personalized Learning Paths

Traditional AI assessments often focus solely on assigning grades or generic feedback. Our solution breaks free from this limitation. We utilize AI for in-depth analysis, pinpointing specific areas where students struggle and for identifying error patterns. This granular understanding allows us to create personalized learning paths for each student, offering targeted resources that address their unique skill gaps. 

Core Innovation

At the heart of our solution lies a patented AI engine focused on strengths-based learning. This approach stands out as a core innovation. Traditional assessments often highlight weaknesses, potentially discouraging students. Our solution flips the script by focusing on a student's strengths. By building upon existing strengths, we can create truly individualized learning journeys that foster a growth mindset and ignite a passion for learning.

AI-Powered Active Replay Technology (ART)

Our proprietary Active Replay Technology (ART) allows teachers, asynchronously, to observe and visualize student thinking.  ART provides a visual heuristic of student engagement, allowing teachers to gauge student confidence and hesitation without direct observation. This tool enhances scalability, enabling a single teacher to manage up to 200 students effectively. We plan to integrate an AI analytics layer to enhance this technology further.

Exam Free Learning

Our ability to track student performance every time they solve problems essentially removes need for periodic summative assessments

Unbiased Assessments and Contextualized Learning

To ensure fairness, our solution employs unbiased initial assessments to establish a neutral baseline for all students.  Furthermore, our AI tailors the learning journey for each student by using a massive pool of assessment items to find the most relevant and unbiased questions that match their current learning  level. AI can adjust the context of a question dynamically to make it more relatable for the student. 

Enhancing Accessibility for All Learners

Recognizing the needs of students with special needs, our system will allow (roadmap) student answers to be submitted as verbal responses. These responses are then converted into transcripts and analyzed by AI. This ensures that all students, regardless of learning style or ability, can benefit from the personalized learning experience.

Catalyzing Positive Impacts in EdTech

Transparency, Auditability, Explainability of our AI: Whatever our AI tech recommends, it needs to be understood by teachers and students so that they can trust the recommendations.  Our explainable AI approach allows us to make AI recommendations transparent, visible and auditable.  These narratives allow us demystify AI decisions, detailing how the AI assesses student performance, tracks activity compared to peers, and determines instructional recommendations. This clarity helps parents and teachers trust and understand the AI’s guidance, and educators to gain valuable data-driven insights to inform instruction and identify at-risk students. 

Elevating Standards: By showcasing the power of AI for personalized learning and ethical data practices, we can push competitors to adopt similar functionalities and prioritize responsible AI use. This can lead to a higher standard for AI-powered learning across the K-12 landscape.

Our functionalities rely on a blend of Artificial Intelligence (AI) subdomains and data analysis techniques. 

AI curated customized learning paths: 

Patented AI Ensemble Model levarages Item Response Theory (IRT RASCH) and Hidden Markov Models (HMM) for in-depth performance analysis and assessment prediction. Thinkster AI using the 2PL Model of IRT/Rasch model, a psychometric model for analyzing categorical data, such as answers to questions, as a function of the trade-off between the respondent's abilities, attitudes, or personality traits. The Thinkster AI continuously updates the difficulty and discrimination probability of each question on daily runs and removes questions with low discrimination indices on a continuous basis. Rasch layer provides recommendations with a confidence score for the concept and overall student ability index.  

Thinkster AI uses Gaussian Hidden Markov model (HMM), a statistical model that is used to describe the evolution of observable events that depend on internal factors, which are not directly observable. We observe learning events and the invisible factor underlying the observation as a “student learning state”. We model every concept as a HMM to determine the learning parameter of student

The Hidden Markov Model measures the learning (concept-wise) of a student over time. Each state represents whether the student knows the concept or not and each transition would represent the learn or forget probability of the student. Thinkster AI computes Student learning states & transitions probabilities to determine if a student has mastered concept. These learning-states provide a  learning sequence with concept assignments based on student curriculum.

Rasch confidence score & HMM learning state and color codified to generate human readable graphical Proficiency Matrix for Students, Parents & Tutors.

Action Replay Technology: It replays the solution steps that the student writes on the app while solving. It can be used to catch error/hesitation/lack of confidence in solving specific steps, giving detailed insights of a student’s proficiency on a particular skill.

Explainable AI

We use Thinkster fine-tuned  Open AI LLM to generate student test narratives from performance data, Rasch recommendations, and HMM Learning states as well Question classification attributes. The narratives are generated for a concept and summarized into Topics to represent student curriculum.

OpenAI LLMs and fine-tune with our in-house content to generate detailed reports. Before sending it to educators/students we do a Sentiment analysis to ensure reports have the correct tone.

We are currently collaborating with Wolfram Alpha to combine their socratic, non-hallucinating AI models and their vast question bank with our personalization and explainable AI models to create a learner intervention solution.  This collaboration will create, near term, a comprehensive learning system that will guarantee learning outcomes for students across the globe and we are very excited about this.

We are also currently working on using Latent Dirichlet allocation (LDA), a Bayesian network for modeling automatically extracted topics in a question bank processed using Thinkster Curriculum. Each Question is then classified using NLP based on trained curriculum and coherence maps.  The classification provides concept mappings, question tags, and other attributes to help in assignments.

We have a working product that has had a transformative impact on student results, with students across the world.

1) We have experience demonstrating impact with our work on instructional support and special needs students

Thinkster has had experience working with Instructional Support students (Title II) as well as special needs students.  We have shown significant and demonstrable results in a pilot program with the South Brunswick School District in New Jersey. 

"Our use of Thinkster Math with students receiving Instructional Support Services at Greenbrook Elementary School has produced exceptional results," said Dr. Gary P. McCartney, Ed. D., Superintendent of Schools for the South Brunswick School District. "Student academic growth, as measured from pre-test through post-test, has increased significantly. When compared to the progress of non-Instructional Support students, students engaged with the Thinkster Math platform experienced a growth rate almost equal to all other grade level students.”  Press Release Link

2) We have provided support for second-language learners

Thinkster is built to support multiple languages to assist with second language learners.  We have already done this with Afrikaans language, a Dutch variant, for use in schools in South Africa very effectively.  This means that –

• Worksheets themselves can be offered in multiple languages.
• Video tutorials associated with worksheets can be offered in multiple languages simultaneously.  Example – a native Spanish speaking student can have video tutorials available in both English and Spanish to help them understand.

3) We have demonstrated proven results for students across multiple countries - please see case-studies, parent reviews, and media reviews for reference.

4) Please see this link for a video demonstrating the product and technology in action. Also, see this video that shows non-hallucinating AI tutor in action.

5) Production use product with Promising Results:

• Our solution, with our existing technology, has undergone successful testing by our in-house educators with expertise in Pre-K-8 learning as well as consumers from over 30 countries and delivered remarkable results.  Since we are continuously innovating, new features are being constantly added to our product stack and can be made available for testing by the Solve team.

• Proven Core Technologies:

  
  

  • Our core components have a proven track record of success:
    • Thousands of students globally have used our existing AI engine and Action Replay Technology for personalized learning paths.
    • User interface operation videos and potentially anonymized student performance data can be provided for these core components.

  
  

Detailed Testing Results Available:

• We're happy to share detailed testing results related to:

  • Question tagging accuracy for difficulty, comprehensibility, and cultural bias.

  • Outlier removal effectiveness.

  • Clustering efficiency for student performance analysis.

  • Correlation analysis for identifying student needs without bias.

  • Effectiveness of question contextualization in improving engagement and understanding.

API Access for Transparent Evaluation:

• We can offer controlled API access to qualified researchers or evaluators interested in a deeper look at the pilot solution's functionalities.

Understanding Diverse Needs:

• Team Composition: We strive for a diverse development team that can identify and address potential biases from different perspectives.

• Community Engagement: We actively engage with educators and families from priority learner backgrounds (public TK-K8, Black/Latino students, and students experiencing poverty) to understand their unique needs and challenges.

• Unbiased Dataset Curation: Our initial question pool is carefully vetted to eliminate cultural or socioeconomic bias. Experts regularly review questions and tagging for fairness.

• Data Diversity Efforts:  We continuously expand our datasets by seeking out student work and assessment data representative of diverse backgrounds, abilities, and learning styles.

• Expert-in-the-Loop:  Subject-matter experts and educators play a crucial role in reviewing AI-generated recommendations and identifying potential bias in question tagging.

Monitoring & Mitigation Throughout Development:

• Bias Auditing: We conduct regular bias audits to identify and mitigate potential biases within the system.

Iterative Improvement: Our solution is designed for continuous refinement. Educator feedback and student impact data directly shape ongoing improvements, ensuring the system adapts to serve learners equitably.

Our core development team consists of 5 full-time staff with expertise in AI, technology, curriculum development, content creation, and product management. Additionally, we collaborate extensively with a network of about about 70 tutor partners who are contractors who provide invaluable feedback. This feedback allows us to continuously improve our models to enhance student learning experiences and mitigate potential bias within the solution.

We are ready for adoption in school environments immediately.  If there is a need to integrate with Google Classrooms or Clever for ease of onboarding students within classrooms, this is the  minimal integration needed to be completed by us.

We have recently been approved as a vendor for Texas schools and are in active discussions with several Education Service Centers within Texas for use and adoption of our technology and product stack at the earliest possible opportunity.

Our overall plan to be ready for large scale adoption among schools focuses on:

• Finalizing key functionalities for user interface optimization.

• Scaling our infrastructure to accommodate a larger student population.

• Implementing robust data security and privacy measures specifically for student information.

• Developing comprehensive training materials for educators.

Our progress is supported by extensive student data from the consumer market showcasing transformative and repeatable results, across a diverse range of student populations.  We have especially seen strong results even among students with special educational needs. We are open to adapting our plan further based on any valuable feedback or support provided by the selection committee.

Affordability and Accessibility Strategies:

• Reduced Pricing for Priority Learners: Public schools serving Black and Latino students, and those experiencing poverty will benefit from significant price reductions through our tiered pricing model.

• Grant Funding and Partnerships: We plan to actively pursue grants and partnerships with educational non-profits to further subsidize access for resource-constrained schools.

• Cloud-Based Delivery: Our secure, cloud-based platform eliminates expensive on-premise infrastructure and minimizes ongoing maintenance costs for schools.

• Affordable Training: Educators will be equipped with comprehensive professional development resources at an accessible price point.

• User-Friendly Interface: We aim to prioritize an intuitive interface with accessibility features, ensuring broad student participation across backgrounds and learning styles.

• Impact-Driven Sustainability: We believe the measurable improvements in student learning outcomes will justify the value proposition for schools, securing long-term adoption.

This multi-faceted approach focuses on breaking down cost barriers and ensuring our solution reaches all learners, fostering equitable access to a powerful educational tool.

The Bill & Melinda Gates Foundation's support would be invaluable in:

• Facilitating Pilot Implementation: Connection with school districts serving priority learners would allow us to gather data on large-scale effectiveness.

• Gathering Diverse Feedback: Pilot participation from diverse school districts would provide invaluable feedback for further improvement.

• Technical Expertise: Guidance and resources from the Foundation's network could accelerate optimization of our solution for broader adoption.