Insight: Machine learning

The concept of machine learning and the benefits it can offer across the education spectrum has become a key talking point among technologists in recent years. But what actually is machine learning and where can it truly have an impact for both teachers and learners? Jo Ruddock finds out

What is machine learning?

Despite becoming an increasingly common phrase, there is still some confusion around machine learning (ML) and how it relates to artificial intelligence (AI). Key to machine learning is data; algorithms are designed to learn from this data and then make a determination or prediction about the subject. In machine learning, computers don’t have to be programmed to complete tasks, it’s about getting them to actually acquire knowledge.

Machine learning is a subset of the much broader world of artificial intelligence, however, AI is more focused on developing a machine that can do something that only a human would normally be able to do.

In the field of education, there are many opportunities for machine learning to make an impact. However, there are also concerns that need to be addressed, not least the vast amounts of data that have to be stored and analysed in order to create effective machine learning algorithms.

Machine learning algorithms are, essentially, used to help teachers provide an education tailored to each student’s strengths, weaknesses, behaviours and habits.
Liz Macfie, Century Tech

Student benefits of machine learning

Firstly, though, let’s look at aspects of education that can benefit from machine learning from a student’s perspective.

Personalised or adaptive learning

A key area in which machine learning is already having an impact is in helping to understand where students are currently at with their learning, and creating personalised plans to help them develop.

As Liz Macfie, head of data science at intelligent teaching and learning platform Century Tech, says: “Machine learning algorithms are, essentially, used to help teachers provide an education tailored to each student’s strengths, weaknesses, behaviours and habits, including the learning content and tests they need in order to maximise their potential.”

Software can identify weaknesses or gaps in a student’s knowledge; it can then determine the most effective learning techniques for each situation and devise a learning plan based on this analysis. At a higher education level, this could have a big impact on student retention rates, identifying those who are struggling or not submitting their work and suggesting steps to improve their engagement levels with a view to reducing dropouts.

Universal access

Machine learning also has a role to play in making classrooms more accessible to students who have learning, visual or hearing impairments, as well as those who speak different languages. Researchers at IBM are using language-processing software developed under the company’s Watson project to make a tool called Content Clarifier that aims to help people with conditions such as autism. It does this by replacing colloquialisms or less common figures of speech with simpler terms, and breaking up lengthy sentences to aid comprehension.

Machine learning could also have an impact on more day-to-day elements of student life, helping neurodiverse students transition to university life by, for example, simplifying the administrative side.

Since an early flush of optimism in the 1950s, smaller subsets of artificial intelligence – first machine learning, then deep learning, a subset of machine learning – have created ever larger disruptions. Infographic courtesy of NVIDIA

Bypassing bias

With machine learning, tests are basically marked blind, leading to claims that it is a useful way to avoid any bias that a tutor may have towards a particular student.

However, contrary to this is the notion that if biased training data was used during the learning cycle, the algorithms themselves could be skewed in their marking.

Google is one company already working to overcome this potential problem and has developed a technique that involves re-weighting a biased data set. Although this increases the time it takes for machines to learn, the results have scored highly on both fairness and accuracy tests.

Smart campus

When combined with a campus’ infrastructure and location-based analytics, machine learning is a crucial element in the smart campus, helping students engage with university life and creating a seamless experience for new starters. For instance, James Clay, head of higher education and students experience at Jisc, told ET last year: “Universities can use the knowledge gained through the gathering of data to ensure that the campus is used effectively to enable an improved student experience, whilst also making efficiencies.”

Phillip Connolly, policy and development manager at Disability UK, described how this machine learning can be implemented for students with additional needs: “Something like a simple navigation app, for a visually impaired person, [that] speaks to them,” could be incredibly useful. “It’s got to be very personalised technology.”

Training is a massive issue for teacher recruitment and retention; if we can use some of the most sophisticated tools to that end, that would be hugely valuable.
Tom Hooper, Third Space Learning

Teachers and machine learning

It’s not just students who could potentially benefit from machine learning; it also has the scope to affect teachers’ lives in a positive way. For example:

Reducing admin hours

Machine learning could impact everything from enrolment to marking tests. Repetitive tasks such as these can be automated, potentially freeing up a teacher’s time. Some multiple choice tests are already being marked by algorithms, and it seems only a matter of time until this can also be the case with more creative tasks. There is also the potential to link this back to the personalised learning mentioned above, with recommendations offered for future learning needs based on test scores.

Professional development

Looking ahead, one key potential area of growth is the use of machine learning to aid teacher training. Tom Hooper, founder and CEO of Third Space Learning, which provides personalised online training, says: “I think some of the most sophisticated tools could take a wider set of data in order to understand the strengths and weaknesses of a teacher, and provide really strong feedback and professional development advice to help with teacher training.

“Training is a massive issue for teacher recruitment and retention; if we can use some of the most sophisticated tools to that end, that would be hugely valuable.”


There’s no denying that anything that can help to streamline the workload of educators and increase the efficiency with which they use their time is critical in education today. However, it’s important not to give too much authority to machine learning without fully understanding it.

Lack of clarity

A lot of questions still surround machine learning, such as who writes the algorithm? What does it mean? Is it actually any good? What’s it picking up on? Are there any biases that we can’t understand? Is that a good thing?

It can be quite hard to understand where machine learning comes from, so it’s important not to fully rely on these results without the all-important human touch of an educator’s input.

Unknown variables

It’s also worth noting that data may not take into account different variables within a student’s life. So, while a machine learning algorithm can analyse data from a series of maths tests, it wouldn’t take into account other variables that can have a significant impact on the data that informs those tests; something as simple as whether the student taking the test had breakfast that morning can impact their performance, for example, but an algorithm will not understand this. If data sets are being trained based on a single strand of a single variable, the risk of coming up with conclusions that don’t understand more significant variables within that student’s overall life is high.


Finally, it should be remembered that machine learning feeds off data – in this case children’s data – and large amounts of it, so institutions should always prioritise data security and confidentiality.

It appears, then, that machine learning offers a number of benefits to both students and educators, saving time and reducing admin for teachers, while potentially also leading to better ongoing training, and offering regular feedback and personalised learning to students. Having said that, relying purely on machine learning data is not the way to go. Teachers know their students better than anyone and there is always the risk of unconscious bias when it comes to data. With that in mind, as in so many areas of edtech, machine learning combined with the human element looks like the way to create the best learning experience for all.

You might also like: How chatbots are changing HE