One of the easiest mistakes in one-on-one education is assuming that because the lesson is private, it is personalized.
I have seen students spend ten weeks working individually with a mentor and need completely different things from the experience. One may arrive with an ambitious research question that needs to be narrowed quickly, while another may be a strong writer who spends weeks moving between topics because nothing has clicked yet. Put both students through the same milestones at the same pace, and they may each receive plenty of individual attention without receiving particularly personalized education.
That is the distinction I think we sometimes miss. One-on-one describes how many people are in the room. Personalization describes what changes because of the person in front of you.
The Ratio Matters, but It Is Not the Whole Story
Strong evidence shows that individual tutoring can help students. According to the Education Endowment Foundation’s review of one-to-one tuition, which draws on 123 studies, students receiving one-to-one support make about five additional months of progress on average [1]. The results become less straightforward, however, when one-to-one tutoring is compared with groups of two or three students, since some studies found that small groups performed just as well or better.
Assigning one student to one teacher creates the opportunity for personalization, but what happens during that hour still matters. A RAND study of personalized-learning schools examined learner profiles, individualized goals, flexible pathways, and different types of student data [2]. One finding stood out to me: 61% of teachers surveyed said they had plenty of student data but needed help turning it into instructional decisions, suggesting that knowing more about a student does not automatically tell an educator what to do differently.
Personalization Sometimes Means Responding Differently to the Same Problem
Two students might arrive at a mentoring session without finishing the work they agreed to do and still need completely different responses. One may normally be reliable and have simply misunderstood the assignment. Another may understand exactly what needs to be done but have avoided the same task for several weeks because it is the hardest part of the project.
A mentor who knows that history is going to approach those conversations differently. One student may need the instructions explained in another way, while the other may need less explanation and firmer accountability around finally working through the part they have been avoiding.
The Search Institute’s Developmental Relationships Framework describes strong relationships with young people as involving support, challenging growth, sharing power, and expanding possibilities [3]. Personalization does not always mean giving a student more help. Sometimes the better response is recognizing that they are ready to do more on their own.
The OECD Learning Compass 2030 also emphasizes student agency and the ability to navigate unfamiliar situations [4]. There is a point where personalization can become too helpful. If every next step is chosen for the student, they have fewer chances to figure out what to do when nobody is guiding them.
Technology Could Give Educators Better Context
When a mentor works with several students, details from different meetings can start to blur together. They may remember that a project has been moving slowly without realizing that the same milestone has now been missed three times, or that the student has already changed direction twice. Looking back across those sessions can change what the mentor decides to spend the next hour discussing.
AI could help with that kind of memory. Before a session, it could pull together what has happened over the past several weeks and point out patterns that might otherwise be easy to overlook. The mentor would still decide whether the student needs more structure, a different goal, or simply some encouragement to move forward, but they would not have to rely entirely on what they happen to remember from the previous meeting.
Both the U.S. Department of Education and UNESCO have argued that people should remain involved in important educational decisions when AI is used [5][6]. In one-on-one learning, that may be one of the more useful jobs for the technology: helping an educator notice what has been happening over time without asking the system to decide what the student needs.
Personalization Is About the Path, Not the Headcount
One-on-one education can be extremely valuable because an educator has time to notice things that would be much harder to spot in a classroom of thirty students. But simply lowering the student-to-teacher ratio doesn't guarantee the experience itself changes.
A student who receives the same process, milestones, interventions, and level of support as everyone else may have an educator entirely to themselves while still following a standardized path.
More one-on-one time can help, but it does not make learning personal on its own. What matters is whether the educator knows the student well enough to realize when an approach that worked for someone else won't work for them.
References
[1] One to One Tuition — Teaching and Learning Toolkit (Education Endowment Foundation): https://educationendowmentfoundation.org.uk/education-evidence/teaching-learning-toolkit/one-to-one-tuition
[2] Informing Progress: Insights on Personalized Learning Implementation and Effects (RAND): https://www.rand.org/content/dam/rand/pubs/research_reports/RR1300/RR1365/RAND_RR1365.pdf
[3] Developmental Relationships Framework (Search Institute): https://searchinstitute.org/resources-hub/developmental-relationships-framework
[4] OECD Learning Compass 2030: https://www.oecd.org/en/data/tools/oecd-learning-compass-2030.html
[5] Artificial Intelligence and the Future of Teaching and Learning (U.S. Department of Education): https://www.ed.gov/media/document/ai-reportpdf-43861.pdf
[6] Guidance for Generative AI in Education and Research (UNESCO): https://www.unesco.org/en/articles/guidance-generative-ai-education-and-research
