Home Artificial Intelligence India’s schoolteachers are drafting higher lesson plans faster, due to a copilot 

India’s schoolteachers are drafting higher lesson plans faster, due to a copilot 

India’s schoolteachers are drafting higher lesson plans faster, due to a copilot 

Shiksha copilot is an element of Project VeLLM, Microsoft Research India’s platform to develop specialized generative AI copilots which might be accessible to everyone from teachers to farmers to small business owners. Shiksha means education in Sanskrit.  

It’s built on Microsoft Azure OpenAI Service and combined with the varsity curriculum and learning objectives. Azure Cognitive Service is used to ingest the text in textbooks including how the content is organized. 

With Shiksha copilot, the researchers hope India’s army of overstretched government schoolteachers could soon get a break – and their students could enjoy livelier and richer lessons.   

From the teacher’s perspective, the largest profit is the time saved preparing for sophistication. Ravindra, who has taught science and math on the Government Higher Primary School Venkatarayanadoddi for 15 years, says he used to spend as much as 40 minutes putting pen to paper to write down a single lesson plan. “Now we will have a brand new lesson inside 10 minutes,” he said.  

On this school of 5 teachers and 69 students, whose parents mostly grow mangoes or rear silkworms for a living, resources may be scarce. So Ravindra modifies the plan as needed. If he doesn’t have materials needed for a suggested activity, he asks Shiksha copilot for an additional idea, specifying what he does have readily available. If a video is simply too long, he asks for a shorter one. He can even modify assignments and ask questions.  

“The old approach to chalk and blackboard is just not sufficient anymore,” said Ravindra. “The time saved by Shiksha, I’m spending with the youngsters.” 

Big class sizes  

Creating lesson plans has all the time been laborious work. A teacher starts with government curricula, normally textbooks, then creates one based on the varsity’s resources, the make-up of learners and the teacher’s own ability and experience. If that’s not hard enough, now try capturing the eye of a generation consumed social media and online videos.  

Here, teachers also grapple with larger class sizes than elsewhere. In Indian primary schools, there may be one teacher to each 33 students versus the world average of 23, in keeping with UNESCO 2020 data. In China, the ratio is 1:16; in Brazil, it’s 1:20 and North America is 1:14.   

In India, as people move from villages to cities, urban class sizes can balloon to between 40 and 80 students, in keeping with the Bengaluru-based Sikshana Foundation.   

Through the years, that has led even individuals with minimal incomes to resort to borrowing money to send their kids to non-public schools, said CEO Prasanna Vadayar.   

Vadayar grew up in Karnataka and moved to the U.S. to pursue a master’s in computer engineering from Texas A&M University before starting a successful software business in Austin, Texas. He returned to India to move Sikshana in 2007 while on a sabbatical. He’s still here. 

Prasanna Vadayar, chief executive officer of Sikshana Foundation, at his Bengaluru office. Photo by Selvaprakash Lakshmanan for Microsoft.

Sikshana Foundation’s mission is to reverse the exodus from government schools by raising the standard of education – permanently. “When people ask me why I joined Sikshana,” said Vadayar, “I tell them ‘to shut it down.’”  

Sikshana is understood for its low-cost, effective interventions that may be adopted by the federal government. It now works in six states across India, covering 50,000 schools and impacting three million students.  

Its Prerana project, for instance, provides opportunities for college kids to be peer leaders and rewards students for small, day by day achievements, from regular attendance to doing well in academics and sports. Students get shiny stars of coloured foil that they’ll collect and safety-pin to their shirts, and colourful stickers denoting lessons learned. Kids get motivated and oldsters can easily see what their child has achieved.  

Prerana was adopted by the Karnataka government in 2018 and rolled out across the state.  

A few years ago, Vadayar noticed how teachers were bogged down preparing for lessons and tried to seek out an answer but gave up, pending the fitting tech. In early 2023, when Microsoft Research India got here knocking with Shiksha copilot, Vadayar remembers considering, “Boom!”   

Generative AI for the masses  

In Microsoft, Sikshana found a partner with the tech solution it was searching for. In Sikshana, Microsoft found an avenue to check the answer in schools and hopefully, mass adoption, possibly beyond India.  

“It is a global problem,” said Vadayar. “The training ecosystem is just not capable of sustain with technology in the category and teachers will not be capable of sustain their learning chops. Kids expect something more.”  

Akshay Nambi and Tanuja Ganu are researchers on Shiksha copilot at Microsoft Research India and have worked closely on other projects. 

Outdoor portrait of a male and female researcher next to each other
(L-R) Akshay Nambi and Tanuja Ganu, researchers at Microsoft Research India, posing for a portrait outside the Sikshana Foundation office in Bengaluru. Photo by Selvaprakash Lakshmanan for Microsoft.

“A 12 months ago, we wanted to take a look at how we could apply GenAI to unravel real-world problems at population scale,” Ganu said. Added Nambi, “Shiksha copilot is a research vehicle for us to grasp how users use them and get user feedback to enhance the general copilot experience.”  

Generative AI tools are built on large language models (LLMs) that synthesize massive amounts of knowledge to generate text, code, images and more. But results may be imperfect. In recent months, researchers have sought to enhance accuracy by adding specific domain knowledge – on this case, the education curriculum of Karnataka state.  

Retrieving information from that knowledge base cuts the danger of errors, said Nambi. That information is then brought into the LLM which generates the lesson plan. The copilot is multimodal, meaning it includes images and video in addition to text, and draws publicly available videos from the web. Finally, Azure OpenAI Service content filters and built-in prompts keep out inappropriate content, for instance, racial or caste-related issues. Through all of it, the teacher is the “expert within the loop,” Nambi said, with students having no direct contact with Shiksha copilot.  

The overwhelming majority of respondents said Shiksha copilot cut their lesson plan preparation from an hour or more to between five and quarter-hour and 90 percent said they only needed to make minor modifications, if any, to the plans generated. Each teacher generated on average three to 4 lesson plans each week.  

“Every class is becoming vigorous”  

For the small cohort of teachers trying out the system, it’s already making a difference.   

The Government Higher Primary School Basavanahalli within the town of Nelamangala on the outskirts of Bengaluru is an L-shaped constructing painted orange and lime green, with a dusty schoolyard where students from grades one to eight assemble and play. The varsity has 13 teachers, 438 students and a mean class size of 30.   

Science and math teacher Mahalakshmi Ashok, who has been testing Shiksha copilot, says it has freed up time for her so as to add activities at school.   

Sitting within the teachers’ meeting room, where the partitions are lined with portraits of Karnataka notables, Mahalakshmi fires up her laptop and opens Shiksha copilot. 

A female teacher at a computer, smiling
Mahalakshmi Ashok, a government schoolteacher and acting headmistress, actively engaged with the Shiksha copilot at Government Higher Primary School in Basavanahalli, Karnataka state, India. Photo by Selvaprakash Lakshmanan for Microsoft.

The primary page is a series of fields with drop-down menus: Select education board, medium (English or Kannada for now, with other local languages to return), class, semester, subject (currently English, Math, Social Science and Science) and topic.  

She picks Science, types “Circulatory System” and selects Duration: 40 minutes. Immediately, Shiksha copilot generates a lesson plan – with the choice to generate a PDF, PowerPoint slides or handout – along with suggested activities, videos and assessment. After each sub-section, there may be a selection of three emojis to rate what’s been generated.  

Previously, Mahalakshmi said, when teaching the cardiovascular system, she may need drawn a diagram of a heart on the blackboard and talked through its function. Recently, she tried a brand new activity suggested by Shiksha copilot. Each student put a finger on their wrist to locate their pulse and measured the beats per minute. They compared results and discussed why some may need a faster or slower heartbeat than others.  

One copilot-suggested activity – a blood-typing lab – was out of the query. But one other activity – taking one’s blood pressure – could be possible. Mahalakshmi thinks she might bring a blood pressure cuff from home next time for the youngsters to check out.  

“Obviously, they like these experiments,” said Mahalakshmi, who has taught for 20 years and can also be the present acting head of college. “Each class is becoming vigorous. The training has grow to be easier.”

A female teacher demonstrating a science experiment to students in a classroom
Mahalakshmi Ashok, a government schoolteacher and acting headmistress, demonstrates a science experiment suggested by the Shiksha copilot at Government Higher Primary School in Basavanahalli, Karnataka state, India. Photo by Selvaprakash Lakshmanan for Microsoft.

At a recent science class on “Separation of Substances,” she brought in small bags of rice, wheat, sand and water as an instance the lesson. The category of 6th graders, wearing all-white uniforms, the women with hair in looped braids, already knew the concepts. They yelled in unison as Mahalakshmi went through the motions, “Handpicking! Winnowing! Sedimentation! Filtration!” and so forth.  

Asked if she had taught this manner before, she said “No,” cocked her head and added, “Because I’m getting more time, no?”  

Beyond lesson plans  

The following phase is to expand the pilot to 100 schools by the top of the tutorial 12 months in March, said Smitha Venkatesh, chief program officer on the Sikshana Foundation. Then starting in April, the team will start curating the best-rated lesson plans, so teachers can modify existing plans as a substitute of getting to create latest ones.  

Smitha, who joined Sikshana 11 years ago after studying and dealing within the U.S., said she has come to grasp the myriad challenges teachers face. 

A female NGO worker standing in a classroom with students in the background
Smitha Venkatesh, chief program officer, Sikshana Foundation, at Government Higher Primary School Venkatarayanadoddi, Kanakapura, Karnataka state, India. Photo by Selvaprakash Lakshmanan for Microsoft.

“There may be the perception that government schools will not be good,” she said. “But we acknowledge that teachers are doing so far more than simply teaching. Not only be certain that students learn, but in addition be certain that they get their uniforms, be certain that they’ve their meals, conduct census, etc. Shiksha copilot might help teachers deliver higher throughout the time constraints.”  

Perhaps in the longer term, Shiksha copilot can even help with other tasks akin to scheduling classes or tracking learning, said Smitha. Perhaps it will possibly also help students who’re struggling.  

“AI is exciting, yes,” she said. “But at the top of the day, can it help the teacher, and may it help the kid?” 


Please enter your comment!
Please enter your name here