Khan Academy is an online learning resource whose learning philosophy is that learning should be:
- Mastery-based to build strong a strong foundation
- Personalized to meet the unique needs of each student
- Interactive and exploratory to encourage creativity and applied learning
These bedrock principles underpin everything we do, and the thousands of letters we get every year from our users are one way we learn how well we are meeting learner needs. However, there is also a strong research foundation underpinning our approach.
1. Effectiveness of Online Learning
In 2010, the US Department of Education conducted a broad meta-analysis  of rigorous published studies that compared online learning with traditional instruction. A key finding from their analysis was that students who took all or part of their course online performed better, on average, than those taking the same course through traditional classroom instruction.
In addition, they found that the positive effects were greater when the online instruction was instructor-directed - i.e., when blended learning was employed - than when the learner worked completely independently.
Khan Academy provides many features to assist learners who use our site, such as the new Learning Dashboard to help users focus their learning, and also provides many coach tools that help teachers guide students’ learning and gain deep insights into their progress. (These features and tools are currently available only when learning math on our site.)
2. Our Learning Principles
Mastery based learning
At its most fundamental, mastery learning simply suggests that students should adequately comprehend a given concept before being expected to understand a more advanced one. Mastery learning is predicated on the belief that all students can learn if provided with conditions appropriate to their needs, and structures its curriculum pacing in terms of target levels of comprehension and achievement, rather than the traditional time-based pacing.
Mastery learning shows strong student learning gains, as shown in Benjamin Bloom’s influential study.  Using Mastery learning, student achievement was a full standard deviation above the achievement attained when traditional learning techniques were used ie: in a mastery learning approach, 84% of the students score above the median score of a traditional classroom, instead of 50%.
Other positive effects have also been shown in research studies - “students in mastery learning programs at all levels showed increased gains in achievement over those in traditional instruction programs”; “students retained what they had learned longer under master learning, both in short-term and long-term studies.”  Another research study found that “mastery learning reduces the academic spread between the slower and faster students without slowing down the faster students.”  There are positive benefits for teachers as well - yet another study found that “teachers who [used] mastery learning...began to feel better about teaching and their roles as teachers.” 
Khan Academy’s learning system is mastery-based and distinguishes several levels of user proficiency. The software guides and encourages learners to do the practice needed to achieve mastery.
Personalized learning within a mastery based model was formally described by Keller  in his Personal System of Instruction (PSI). The defining characteristics of his system included self-pacing - permitting each student to progress through material at a pace suited to him/her; mastery based learning of sequential units; frequent and repeated assessments with immediate feedback; and support from coaches, who could either be teachers or advanced peers, to provide tutoring and guidance. In this approach, the teacher’s role changes from primarily that of a lecturer to a guide who provides highly targeted tutoring and assistance based on the individual learning needs of each student.
These characteristics of Keller’s personalized system of instruction are key elements of Khan Academy’s learning philosophy.
A meta-analysis  of 75 comparative studies of Keller’s system showed that personalized learning generally “produces superior student achievement, less variation in achievement, and higher student ratings in college courses.” The analysis also showed personalized learning produced superior results in a variety of course settings, and with a variety of research methodologies.
Another meta-analysis  of seven different studies concluded that “[Keller’s PSI] is effective in fostering improved subject matter mastery over more conventional instructional approaches. This is true regardless of whether the synthesis is quantitative or narrative.”
In addition, these studies showed that students in courses using personalized learning achieved higher test scores with less instructional time, and enjoyed their courses more, compared with the traditional lecture method.
Personalized learning also stresses the importance of rapid feedback for learners. The effect of feedback on student learning gains was shown by John Hattie  who conducted 800 meta-analyses encompassing more than 50,000 studies and millions of students, and concluded that timely feedback had the largest effect size on learning of any intervention studied. Studies by Epstein et al also show the importance of rapid feedback. 
Khan Academy’s online learning system supports students learning at their own pace. Our online problem sets provide immediate feedback on whether students have worked the problem correctly, as well as progressive hints, so students do not get stuck, but are able to continue to make progress. The new Learning Dashboard helps guide student learning. In addition, our in-depth Coach Reports provide deep insights for teachers on exactly where their students need help.
Together with features like our Coach Recommendations, teachers have the information they need to provide precisely targeted help. (At present, these features are available only for our math content.)
Interactive and Exploratory Learning
The third pillar of Khan Academy’s learning approach is that learning should be interactive and exploratory.
This means that learners are actively engaged with the material they are learning and interact and collaborate with their peers, rather than passively absorbing information.
This implies problem solving and project based learning, as well as peer collaboration, together with the timely feedback and targeted assistance from coaches already discussed as part of Personalized Learning.
Problem and Project Based Learning: In problem-based learning, learners work on solving real world problems, while project based learning involves a complex task focused on a driving question involving central concepts of one or more disciplines. These learning models are student-driven and involve inquiry, investigation, knowledge building and decision-making, while teachers serve as coaches and facilitators. [11, 12] The goals of these learning models are that students learn or reinforce abstract academic content by applying it to real world situations and problems.
When implemented well, studies show that these approaches increase long term knowledge retention, and improve collaboration and problem-solving skills as well as student attitudes to learning. [13,14]
Khan Academy encourages using the classroom time that is freed up by adoption of mastery based, self-paced learning for project based learning. To assist teachers, we have created step-by-step guides for some interesting projects that enable teachers to create hands-on learning platforms (eg: the popular robot projects Bit-zee, Spider, and Spout) that can be used to teach a variety of science and math principles.
Peer Collaboration: In collaborative learning, students work together in small groups to master academic material. Studies show that collaborative learning methods show “statistically significant and substantially greater” student learning gains than traditional classroom approaches  and that student relationships and ability to work together, as well as positive attitudes towards learning, improve along with student achievement. 
Peer tutoring, where students who have mastered a concept assist their peers, has also been shown to help both tutors and tutees.  Tutees “outperformed control students” on achievement, while also “developing more positive attitudes toward the subject matter”. Tutors benefit as well - tutors also “gained a better understanding of, and developed more positive attitudes toward” the material covered.
A key part of Khan Academy’s learning vision is to incorporate project and problem based learning, as well as peer tutoring and collaboration, and using Khan Academy to support self-paced learning frees up class time for this collaborative and exploratory learning.
3. Khan Academy Research and Platform Enhancements
In addition to the development work that supports our foundational learning philosophy, we continuously incorporate improvements into our platform based on our original research. With millions of problems worked every day, our researchers have access to one of the largest datasets available in education. Areas of active Khan Academy research include:
- Psychology - In collaboration with external researchers from Stanford, Berkeley, and elsewhere, we have studied the effects of “growth mindset”  related messages on the Khan Academy website and found that users who viewed the messages showed increased success and motivation, as measured by skills mastered, problems attempted, and return visits. Building on that success, we are creating experiments that leverage other psychological aspects of learning, including the “explanation effect” and development of student “grit”.
- Assessment - We have implemented the research-supported best practice of evidence-centered design in our creation of assessments. We have extended the Item Response Theory framework with novel improvements to make our assessments as efficient and accurate as possible.
- Content analytics - We continually perform randomized controlled experiments using variations of our content and its organization, and measure the effectiveness of each variation. Often, the variations are based on specific findings from the cognitive sciences on how people learn.
- Educational Data Mining - By using machine learning techniques, we attempt to optimize our automated learning recommendations to students. While using sophisticated empirical techniques, we design algorithms and user experiences that take maximal advantage of effective learning strategies such as distributed and interleaved practice. 
1. U.S. Department of Education. (2010): Evaluation of Evidence-Based Practices in Online Learning: A Meta-Analysis and Review of Online Learning Studies. http://www2.ed.gov/rschstat/eval/tech/evidence-based-practices/finalreport.pdf
2. Benjamin S. Bloom (1984): The 2 Sigma Problem: The Search for Methods of Group Instruction as Effective as One-to-One Tutoring, Educational Researcher Vol. 13, No. 6 (Jun. - Jul., 1984), pp. 4-16
3. Guskey, T. R., & Gates, S. (1986): Synthesis of research on the effects of mastery learning in elementary and secondary classrooms. Educational Leadership, 43, 73-80.
4. Daniel U. Levine (1985): Improving student achievement through mastery learning programs, Jossey-Bass.
5. Davis, D., & Sorrell, J. (1995): Mastery learning in public schools. Educational Psychology Interactive.
6. Keller, F. S. (1968): Goodbye teacher…Journal of Applied Behavior Analysis 1, 79-89.
7. JA Kulik, CLC Kulik, PA Cohen (1979): A meta-analysis of outcome studies of Keller's personalized system of instruction, American Psychologist.
8. Pascarella, Ernest T. & Terenzini, Patrick T. (1991): How College Affects Students, Jossey Bass.
9. Hattie, J. (2008): Visible Learning: A synthesis of over 800 meta-analyses relating to achievement, Routledge
10. Epstein, M., Lazarus, A., Calvano, T., Matthews, K., Hendel, R., Epstin, B., Brosvic, G., (2002): Immediate Feedback Assessment Technique Promotes Learning and Corrects Inaccurate First Responses. The Psychological Record, 52, 187-201
11. Barron, B., & Darling-Hammond, L. (2008): Teaching for meaningful learning: A review of research on inquiry-based and cooperative learning (PDF).
12. Thomas, J. W. (2000): A review of research on project-based learning (PDF).
13. Strobel, J., & van Barneveld, A. (2009): When is PBL more effective? A meta-synthesis of meta-analyses comparing PBL to conventional classrooms. The Interdisciplinary Journal of Problem-Based Learning, 3(1).
14. Mergendoller, J. R., Maxwell, N. L. , & Bellisimo, Y. (2006): The Effectiveness of Problem-Based Instruction: A Comparative Study of Instructional Methods and Student Characteristics. Interdisciplinary Journal of Problem-based Learning, 1(2).
15. Terenzini, P. T., Cabrera, A. F., Colbeck, C. L., Parente, J. M., & Bjorkland, S. A. (2001): Collaborative learning vs. lecture/discussion: Students' reported learning gains. Journal of Engineering Education, 90(1), 123-130.
16. Slavin, R. (1991). Synthesis of research of cooperative learning. Educational Leadership 48(5), 71-82.
17. Peter A Cohen, James A. Kulik and Chen-Lin C. Kulik: Educational Outcomes of Tutoring: A meta-analysis of Findings, Am Educ Res J 1982 19: 237
18. Carol Dweck (2006): Mindset - The New Psychology of Success Random House
19. John Dunlosky, Katherine A. Rawson, Elizabeth J. Marsh, Mitchell J. Nathan, and Daniel T. Willingham (2013): Improving Students’ Learning With Effective Learning Techniques: Promising Directions From Cognitive and Educational Psychology, Psychological Science in the Public Interest 14(1) 4–58