“Charting the Path Forward for Tutoring Based in Evidence and Practice” was the first time Accelerate.us hosted an in-person gathering. Held April 26-27th in Arlington, Virginia, this invite-only event was extended to 31 grantees and provided valuable insights, fresh ideas, and networking prospects within the tutoring community and K-12 Education System.
Investing in the K-12 Education System
John Failla, Founder and CEO at Pearl, and Nate Casey, Chief Strategy Officer at Pearl, spent the day with the Acclerate.us Community of Practice listening to panel discussions, participating in breakout sessions, and connecting with fellow grantees. The dedication to incorporating tutoring as a fundamental component of the K-12 education system in the upcoming years has never been more apparent.
Attendees were a wonderful blend of K-12 Education professionals ranging from classroom educators, researchers, policy experts, to EdTech innovators. As panel discussions unfolded and workshops began, everyone contributed their unique knowledge and valuable experiences, making this a truly inspirational event.
Accelerate.us Conference Key Takeaways
Accelerate’s CEP, Kevin Huffman initiated the conference with an optimistic, yet grounded, perspective on the past two years and the challenges ahead. The room was pervaded by a palpable sense of urgency to demonstrate the efficacy of our approach before the ARP ESSER funding expires in September 2024.
The opening panel featured district leaders from Baltimore and Washington DC, who shared intriguing insights into the complexities of implementing tutoring programs within the K-12 education system. In particular, Arthur Mola, Principal of the Cardoza Education Campus captivated the audience through his discussion on the importance of having a dedicated tutoring lead and the necessary steps to foster genuine collaboration with external support.
Implementing Tutoring Programs
Having a dedicated tutoring coordinator on your campus greatly streamlines the integration of tutoring into the scheduling framework
Engaging kids in the data that shows their personal academic growth can be a powerful motivating factor.
Rolling out MTSS anew is hard, but can become a strong catalyst for weaving tutoring more thoughtfully into the K-12 experience.
Future of Tech
ChatGPT and AI writing software is not just a clear game changer, it has the opportunity to impact education in many ways, including professional development, assessment, and driving deeper individualized learning pathways.
Although new technologies have a tremendous amount of promise, the “relationship” between student and tutor is still a critical aspect to learning.
New technology is focused on driving down the cost of individualized learning which may soon allow for a single teacher to teach a classroom of students with individual learning gap data feeding “in real time” to the teacher.
Ensuring Outcomes Through Policy
Every case study examined (Arkansas, Tennessee, and Texas) took a different approach to leveraging policy as a way to move quickly on tutoring.
In many instances, the most meaningful policies are the ones that unblock heavy barriers to entry, like streamlining criminal background checks or incentivizing tutors with better pay.
There is a great opportunity to incentivize programs that adhere to strict evidence-based approaches to delivering tutoring (i.e. – TQIS as defined by the NSSA)
Sustainability and The Future
Increased emphasis on personalized learning: With advances in educational technology and the availability of data-driven insights, more school districts are likely to adopt personalized learning models that cater to the unique needs and preferences of individual students. This may include a greater use of adaptive learning platforms, virtual tutors, and other tools that enable students to learn at their own pace and according to their own learning styles.
Greater collaboration between schools and external stakeholders: To address the increasing demand for high-quality tutoring services, K-12 education systems and school districts may partner with external stakeholders such as community organizations, non-profits, and private companies to provide additional resources and support. This may involve the development of public-private partnerships, grant programs, and other initiatives that bring together various stakeholders to improve access and equity in education.
Enhanced infrastructure to support remote learning: In light of the COVID-19 pandemic, many school districts have already invested in upgrading their infrastructure to support remote learning. This trend is likely to continue in the next five years, with more schools adopting blended learning models that combine traditional classroom instruction with online and remote learning opportunities. This may require investments in new hardware and software, as well as training and professional development for teachers and staff to support effective remote instruction.
Accelerate actively builds the country’s knowledge of tools and practices that significantly advance student learning. [They] do this by bridging the gaps between research, communities, and systems to ensure the rapid and widespread adoption of proven strategies.
Since ChatGPT was released to the public via OpenAI the education community has been in constant chatter about its implications for learning. A large language model is a machine learning system that uses vast amounts of data to generate human-like language and perform natural language processing tasks such as text generation, translation, summarization, and sentiment analysis.
If you have not experienced ChatGPT you should. It’s definitely an exciting next generation in how we think about searching on the internet. For almost a quarter of a century Google has been the cornerstone of where we find information online. Google presents relevant links based on search. With ChatGPT information is presented in natural language and feels much more like a chat with someone very knowledgeable about the question you ask.
Although the conversation feels fairly natural and the answers are cogent, they are not always correct. That being said, if you have a decent understanding of a subject you will probably find that the chat feels authentic and most of the time pretty accurate.
When the AI was released everyone at Pearl was interested in jumping quickly and trying to understand the implications for tutoring. We spend a lot of time thinking about equity and imagine a world where anyone that needs a tutor could have a tutor. None of us is convinced that AI could replace a human, but I do think that AI has the capacity to make someone better at tutoring or give more opportunities for people to teach.
Here are some interesting implications for the tutoring industry:
Without compromising any person’s identity a tutor admin can give a list of characteristics of a learner and then ask to see corresponding characteristics of a tutor. Below is an example of a question and ChatGPT’s answer:
“Based on the research, what are 5 strong characteristics to look for in a tutor that would match well with a 9 year old Asian dyslexic child coming from a poor socioeconomic background learning in English whose first language was Chinese?”
“Great question! Here are five strong characteristics to look for in a tutor that could match with an Asian 9-year-old dyslexic child coming from a poor socioeconomic background learning in English whose first language was Chinese:
Patience: A tutor who is patient can help the child feel comfortable and confident as they learn. Dyslexia can be frustrating and challenging for a child, especially when they are also trying to learn a new language. A patient tutor can help the child to feel supported and encouraged.
Multisensory approach: Dyslexic children can benefit from a tutor who uses a multisensory approach to learning. This means that the tutor can use different senses, such as sight, sound, and touch, to help the child learn. For example, the tutor could use visual aids, like pictures or diagrams, or use physical activities to help the child understand the material.
Experience working with dyslexic children: It’s important to find a tutor who has experience working with dyslexic children. This experience can help the tutor understand the unique challenges that dyslexic children face and how to tailor their teaching methods to meet the child’s needs.
Cultural sensitivity: A tutor who is culturally sensitive can help the child feel understood and respected. It’s important to find a tutor who is familiar with the child’s cultural background and can be sensitive to their unique needs and experiences.
Flexibility: Finally, it’s important to find a tutor who is flexible and can adapt to the child’s needs. Dyslexic children can learn at different paces and in different ways, so the tutor should be willing to adjust their teaching methods to meet the child’s individual needs. Additionally, the tutor should be flexible in terms of scheduling and availability to ensure that the child can receive consistent support.”
Once ChatGPT or some version of this type of AI is available, an admin could ask AI to construct a calendar based on the data for two tutors. Here is an interesting example of ChatGPT’s response to a basic scheduling prompt.
I am a tutor that needs to meet with my student three times a week for 30 minutes each time. I would like to know the dates I can meet with that student in school based on a typical school calendar.
The specific school calendar will depend on the school district and location. However, in general, most schools in the United States operate on a traditional academic calendar, with two semesters per year (fall and spring) and breaks for holidays and summer vacation.
To determine the possible meeting dates, you can start by looking up the school district’s academic calendar online or contacting the school directly to get a copy of the calendar.
Assuming a typical school calendar, here is a general example of possible meeting dates for a tutor meeting three times a week for 30 minutes each time:
Fall Semester: September to December (approximately 16 weeks)
Possible meeting dates:
Monday, Wednesday, Friday: September 6th to December 16th (excluding holidays)
Tuesday, Thursday, Friday: September 7th to December 16th (excluding holidays)
Spring Semester: January to May (approximately 16 weeks)
Possible meeting dates:
Monday, Wednesday, Friday: January 10th to May 20th (excluding holidays)
Tuesday, Thursday, Friday: January 11th to May 20th (excluding holidays)
Keep in mind that these dates are approximate and may vary depending on the specific school calendar. It’s important to confirm the exact school calendar with the school or district to ensure that the meeting dates are accurate.”
We’re starting to get into the interesting part now. Obviously, it is imperative that the teacher’s knowledge confirms, in some way, whether the ChatGPT response is helpful. As an example, ChatGPT can help tutors contextualize or better consider a student’s learning style or interests in order to support their tutoring. Let’s see if ChatGPT builds some math problems based on two factors (age and affinity) based on my son’s interest in trucks and construction equipment.
My son is 12 years old and has a strong affinity for trucks and construction equipment. Based on his age and interests please produce 3 great ways to teach him long division.
If your son has a strong interest in trucks and construction equipment, incorporating these interests into his long division lessons may help him stay engaged and motivated. Here are three great ways to teach him long division:
Use visual aids: Use pictures or diagrams of trucks and construction equipment to explain the steps of long division. For example, you could use a picture of a dump truck to represent the dividend and use smaller pictures of construction equipment, such as bulldozers or excavators, to represent the divisor and quotient. This can help your son visualize the process and make it more concrete and meaningful.
Turn division into a game: Create a game that uses trucks and construction equipment to teach long division. For example, you could create a game where your son has to use a dump truck to carry a certain number of building materials and use excavators to divide the materials into equal groups. Each time he successfully divides the materials, he can move to the next level of the game. This can make the learning process more fun and interactive.
Use real-life examples: Find examples of how trucks and construction equipment are used in real-life scenarios that involve division. For example, you could use examples of construction projects where trucks have to haul and divide large quantities of materials. This can help your son see the practical applications of division and make it more relevant to his interests.
Remember, every child learns differently, so it’s important to find an approach that works best for your son. Incorporating his interests and making learning fun can go a long way in helping him stay motivated and engaged in the learning process.
Assessment is the final interesting implication for tutoring. At Pearl, we believe the key to tutoring is quick formative assessments that verify student learning and identify areas where knowledge or understanding might be lacking. Once again, this is a great way to leverage several variables in a question. The following are two examples that you may find interesting:
Please give me three formative assessment questions that will help me validate that my 9 year old student understands long division.
Can you explain the steps involved in long division and how they relate to each other?
Can you solve a long division problem independently and explain how you arrived at your answer?
Can you apply long division to solve real-world problems and explain how you arrived at your answer?
Please provide me with 3 problems to solve that would assess my 9 year old’s understanding of the basics of long division.
84 ÷ 6 = ?
145 ÷ 5 = ?
232 ÷ 8 = ?
Note: These problems are designed to test the student’s understanding of the basic steps of long division. When assessing your student’s responses, make sure to check that they have correctly used the divisor, dividend, quotient, and remainder, and that their final answer is correct. It’s important to provide feedback on their work and address any areas where they may need additional support or practice.
When tutoring programs consider using ChatGPT or any of the other emerging large language model AI tools, they must validate the tools carefully. It is possible to speed up certain processes, but there is a cost associated with it. There is a possibility that AI is extrapolating biases that are not aligned with your ethics. Additionally, tutors should cultivate value that reflects their approach to tutoring and not rely too heavily on computers. Relationships are at the heart of tutoring, so it’s important to foster connections and care throughout the process. Nevertheless, keeping up with the changing world is essential. There is no doubt that our students will benefit from this new tech; therefore, as educators to the next generation, we need to keep our eyes on the ball.
Now, for kicks, let’s let ChatGPT have the final word ?
What are the implications of Large Language Machine Learning Models in Tutoring
Large language machine learning models have several potential implications in tutoring, including:
Personalized Learning: These models can help provide more personalized learning experiences for students by tailoring the content and pace of the instruction to their individual needs, interests, and learning styles.
Immediate Feedback: The models can provide immediate feedback to students on their work and progress, helping them identify areas where they need additional practice or support. This can help students stay engaged and motivated and can also help teachers more effectively monitor student progress.
Automated Scoring and Grading: These models can help automate the grading and scoring of student work, freeing up time for teachers to focus on providing individualized instruction and support to students.
Improved Accessibility: These models can help make tutoring more accessible to a wider range of students, including those with disabilities or who speak different languages.
Improved Instructional Design: These models can help identify patterns in student learning and suggest improvements to instructional design and content, helping teachers to more effectively support student learning and improve outcomes.
However, there are also some potential challenges and limitations to the use of large language machine learning models in tutoring, including issues of privacy and data security, the potential for bias in the algorithms used to analyze and evaluate student work, and the need for teachers to be trained in how to effectively use and interpret the results of these models.
The many variations across hundreds of tutoring RFPs
Each year our public school districts issue hundreds of requests for proposals (RFPs) for high-impact tutoring support, but there is no standard approach to collecting feedback and assessments of the tutoring. Many districts require their tutoring providers to administer academic assessments, while others do not. Here at Pearl, our partners typically follow one of these 4 approaches to assessments:
The tutoring provider uses a specific assessment tool (like iReady or NWEA MAP) to assess students at all stages of a student’s tutoring experience.
The tutoring provider assesses each student’s progress (choosing their preferred benchmark, formative, and other assessment types) with no specific assessment tool requirement other than the assessment questions used must meet the state’s ESSA standards.
The tutoring provider supports students but is not required to administer the assessments. In this case, the district runs the assessments and advises the tutoring provider accordingly.
A state-based program serving multiple districts is required to use different assessment tools depending on the requirements of each district they serve. In these cases, the state agency is also left with the complicated task of comparing district progress with varied types of outputs and insights from multiple assessment tools.
The Pearl platform provides the technical scaffolding for state and district tutoring programs, nonprofits, and enterprise-level tutoring companies. For each of these different client types, we take an agnostic approach to how academic progress is assessed. Sometimes our partners have their own assessments, like district-funded programs, and other times they simply follow the requirements defined in the contracts they serve, such as outside tutoring companies responding to RFPs.
In a “best case” scenario tutoring providers would have:
A diagnostic or benchmark assessment from the student’s school (or district) to inform gaps in learning and insights into how that student learns most effectively
A viable feedback loop with a student’s school teacher(s)
A formative assessment tool that:
Meets the applicable state ESSA standards
Aligns with the school’s curriculum
Is agile enough to be used on the fly
Provides a way to regularly measure student academic progress
Informs clear opportunities for tutors to employ an individualized learning pathway for their students
A final “end of tutoring” assessment to measure progress against the benchmark test
If more districts had the resources to increase their level of system interoperability, the above “best case” assessment pathway would be more widely viable. It is common for even sophisticated districts to lack an elegant method for merging assessment data from providers with SIS student records. In addition, tutors may be unable to access useful academic-related data because of security concerns and requirements.
In the absence of a standard assessment methodology, a host of problems arise. To determine learning gaps, tutors often need to conduct several sessions without a baseline. If a tutor does not use a formative assessment approach aligned with the student’s starting point, he or she may be left guessing about the student’s understanding gaps. Also with so many assessment approach variations, it is difficult to compare programs district-to-district or state-to-state.
To develop a successful tutoring program, four key data categories are needed (attendance, dosage, academic progress measurement, and SEL insights). In the absence of academic progress that relates to individual students, tutor providers are unable to measure the true picture of progress or show clear ROI to stakeholders. With the final ESSER cliff fast approaching it is critical that programs robustly measure efficacy. Only programs that demonstrate evidence of effectiveness will be sustainably funded in the future.
Various tutoring approaches are being implemented across the US to address learning loss. The most common are high-impact tutoring, on-demand tutoring (sometimes referred to as homework help), and a hybrid model in which certain students in a program receive scheduled high-impact tutoring and the general student population has access to opt-in on-demand academic support.
High-Impact-Tutoring (also sometimes referred to as High-Dosage-Tutoring)
A high-impact-tutoring model optimally leverages the NSSA’s “Tutor Quality Improvement System (TQIS) and include “substantial time each week of required tutoring, sustained and strong relationships between students and their tutors, close monitoring of student knowledge and skills, alignment with school curriculum, and oversight of tutors to assure quality interactions.” Although this model has strong examples that are grant based and self administered at the state or district level, it can also be outsourced to a tutoring vendor.
On-Demand Tutoring (also sometimes referred to as Homework Help)
24/7 access to tutors via chat and whiteboard across all key subjects is the most popular style of on-demand tutoring. On-demand homework help (or as-needed academic assistance) allows students to meet with tutors as needed, often in a short time frame (like 10-15 minutes) to discuss something they do not understand – usually related to preparing for a test or completing homework (often due the next day). Curriculum or learning management systems are not usually required in this model, although tools aligned with specific approaches may be used. Most often, this model is outsourced to an outside tutoring company, but some examples of this model are found in community tutoring partnerships, such as Dakota Dreams in South Dakota.
Combining High-Impact-Tutoring and On-demand Tutoring (Hybrid)
On-demand homework help and high-impact tutoring are combined in the final model. In the early stages of exploring this hybrid model, vendors are scheduling out high-impact tutoring as usual but also giving students the option to receive on-demand help in other subjects online. Furthermore, on-demand access is being explored across an entire district in the same environment where high-impact tutoring is taking place.
Deciding on Your Approach
An effective tutoring program should not be developed based on a best guess or someone’s perception of what is possible. The design of a successful tutoring program should be based on logic. According to the National Student Support Accelerator’s (NSSA) toolkit for program design, any tutoring program should outline: needs, inputs, actions, outputs, and impact.
Needs: what needs the program address?
Inputs: what goes into the program?
Actions: what actions does the program take?
Outputs: what happens as a result of those actions?
Impact: What are the benefits of participating in the program?
Based on the logic model and clear definition of each component of the model, stakeholders can design a program for the desired outcome.
During the 2022 holiday, Southwest’s antiquated tool, SkySolver, and various internal processes collapsed. It was not a matter of broken software that caused this unforgettable meltdown. A strategy to deal with a disruptive storm was undermined by a lack of modern automation and poor user experience. It was simply not possible for Southwest to handle the sheer volume of changes and coordination required without a modern digital infrastructure. In addition, Southwest lacked a central hub for gathering resources, regrouping, and finding a way out of the chaos. These problems are not unique to Southwest. Companies across all industries should evaluate their dependence on homegrown systems, especially those built on antiquated mainframe computers.
Reflections on the Disaster
Our team closely monitored Southwest’s situation as well as the subsequent collapse of the FAA’s ability to keep the entire US flying. We arrived at the conclusion that our platform, in many ways, serves a similar function to an airline. In several meetings, we reflected on the fact that what we do as a company (like an airline) is provide our partners with a reliable set of solutions that empower them to tutor thousands of students each day.
Witnessing a Fortune 500 company collapse due to inadequate logistical automation was humbling, but also incredibly insightful. We talked about how the heart of our mission is simply making sure tutors and students can be together (attendance). Consistently bringing people together is the crucial first step in preventing a catastrophic and possibly generational failure in academic human rights.
How is the Airline Industry Like Tutor Program Management?
Picking up a loved one
Service being delivered
It takes a great deal of coordination and effort to run a massive tutoring initiative. Administrators (operations in a central hub) coordinate school schedules, deal with HR issues, and help parents(like a family member waiting for an arrival) who are confused or unhappy for various reasons. Each session “takes flight” under the direction of these administrators, ensuring that every student receives the support they need. Tutors (like pilots) show up in rooms without students, looking for administrators or tutor coaches to fill the void of a student’s absence.
Teachers (like ground crew) keep an eye on their clocks to make sure students receiving tutoring are in the right place at the right time. There are times when students(like travelers) need to miss class or navigate to another tutoring venue which may be outside their regular schedule. In order for online learners to login at the right time, they need to be reminded or sent messages. There are times when tutors have to be substituted because of cancellations. While all of this is happening, the world and policymakers are waiting for results. Administrators are charged with the difficult task of making sure consistent standard data is collected so that the overall health of the program can be evaluated for its impact and for its many opportunities for improvement.
Behind all of the many wonderful stories about tutoring as a source to promote academic progress are a group of people dedicated to managing the overwhelming logistical components.
Running a tutoring program that leverages thousands of tutors serving tens of thousands of students requires:
Designing a logic model with clear definitions and alignment on Needs, Inputs, Actions, Outputs, and Impact
Thoughtfully matching students with the best tutor – leveraging logistical data and evidence based data points that have proven better outcomes
Familiarizing tutors with the technology used to coordinate logistics, communicate efficiently, teach with clarity, run assessment, and collect the right data (attendance, dosage, academic progress, and social-emotional-learning insights)
On a daily basis, making sure every scheduled student spends time with their tutor
Leveraging 360 feedback loops to make on-the-fly changes to improve program design
Reflecting on outcomes to optimize for success in the next cycle
One of the most critical industries in this country failed millions of users because of legacy systems and poor operational design. It is imperative that we do not let the tutoring industry, which is just as vital to our country’s future, suffer the same fate.
The heroes that run our nation’s massive tutoring programs deserve:
Highly secured cloud-based software for efficient program implementation (rostering, matching, seamless communications, data collection, and much more)
Modern technology that evolves in real time (with partner collaboration) affording administrators the technical scaffolding to constantly make their lives easier
In-app chat to ensure all end-users that experience technical difficulties have a human ready to help at a moments notice
Most importantly, a technical partner that listens every day, and works tirelessly with partners to explore ways to improve: automation, user experiences, and student achievement