A New York school district made global headlines this week for plans to trial a humanoid robot named “Sally” in high school classrooms.
Sally has lifelike silicone skin, long brown hair, and the ability to move her arms, head, and face while seated. According to the manufacturer, she uses “natural conversation, facial expressions and real-time interaction”.
She is programmed to be a stationary classroom teaching assistant and would work as a chatbot to answer student and teacher questions. She would also be available to students 24/7 as a digital avatar online.
Following concerns about student privacy and data storage and tracking, the project has been paused.
But how far away are we from having human-like robots or AI-enhanced avatars as routine features of the school classroom?
Robots are not new to classrooms
Sally is not the first robot to be proposed for classrooms.
Non-humanlike robots (such as “Nao”) have been trialled in classrooms before. For example, they have been used to help students learning languages in Japan, South Korea, Europe and East Asia. They have also been used in technology classes when teaching skills such as coding.
Tools such as Alexa, which responds to voice commands, have also been trialled in primary classrooms in New Zealand. Here, results varied by age. Younger children had difficulty asking questions in the right format to get the answer they were seeking. Older students tested the limits of the technology, testing social interactions like joke telling or seeking unusual responses.
Teachers are also increasingly turning to generative AI for administrative tasks to save time, while chatbots are now regularly used in schools in Australia. For example, South Australia and New South Wales have introduced their own generative AI tool for students and teachers.
This growing day-to-day familiarity is likely to lower the barrier to more advanced, human-like technologies entering classrooms.
Here to help
Research on AI-based avatars (which chat like a human but only exist online) found they can have a positive effect on learning.
They can provide on-demand academic support by individualising learning. For example, patiently going over a maths problem, step-by-step.
They can also help self regulation by teaching strategies and repeatedly encouraging students to apply these over time. For example, “before you start this equation, take a deep breath”.
Students have also reported experiencing less pressure and anxiety when interacting with AI tutors, which are often perceived as less judgemental than teachers.
But we still need real teachers
A robot can simulate warmth, deliver curriculum flawlessly, and never get tired or need a break.
However, evidence indicates the most effective role for these technologies when they are used alongside teachers, rather than replacing them.
For example, a 2026 meta-analysis of 146 studies on social robots (robots that have a physical presence in the classroom) used to teach cognitive skills, found they could improve learning outcomes, especially when they could personalise responses. But this was only when a teacher was present.
But robots used as teacher substitutes could not manage the complexity of the classroom content, respond to spontaneous student behaviours or handle unforeseen occurrences.
Research also tells us children and young people learn best when they are engaged not just cognitively, but with other people.
They need interactions that build trust and instil confidence they are seen, known and cared about by their teachers and peers in human ways that make sense to all involved.
Data, privacy and beyond
There are safety concerns as well with this technology.
As a robot interacts with a child, the interaction is recorded. This allows the robot to provide targeted and personalised responses. When programmed with generative AI, the data may be accumulated and used to inform future interactions with other students.
But this creates risks around data privacy, storage and tracking.
It also leads to other questions about who programs and then controls the programs for the robots, and to what end? What happens if the program or data is hacked?
Can kids learn empathy from a robot?
Teachers are in a significant position of trust, in charge of vulnerable children and young people. There are legislative frameworks and teacher registration boards that regulate teachers’ professional behaviour but does not extend to robots.
Teachers are obliged to identify, protect and foster students’ best interests. This requires them to make nuanced judgements about individual learning needs, behaviour and wellbeing.
This can be difficult to capture in fixed rules or instructions, or robot programming.
As researchers have noted, this also raises concerns about children’s development if they are interacting with simulated rather than genuine care and empathy.
What’s next?
We are not going to have human-like robots in classrooms tomorrow. But the developments out of New York show how they might be a serious option for the future – particularly amid ongoing concerns about teacher shortages and workloads.
If we are going to move further in this direction, there are tough questions to answer. We will need teachers present to make sure they are being used in ways that help and protect students as they interact and learn.
The post “A lifelike robot called ‘Sally’ was going to teach in a US school. It’s not a weird as it sounds” by Louise Starkey, Professor of Education, University of Tasmania was published on 07/29/2026 by theconversation.com