Headteachers have criticised plans to use artificial intelligence to help identify failing schools under a government scheme.

The National Association of Head Teachers says that the plan for using algorithms will not help the effective inspection of schools and Ofsted inspections should not be affected or influenced by the use of data.

The call comes after it was revealed that there are plans to use a data science unit to work alongside Ofsted in England to help prioritise which schools should be inspected.

The unit is owned partly by the government and an algorithm has been created to help rate a school using machine learning, otherwise known as artificial intelligence.

A spokesman for the NAHT said: “We need to move from data-led approaches to school inspection since it’s important the process is transparent and schools can learn and understand from any assessment.”

‘The inspection system will treat leaders and teachers fairly’

The spokesman added: “Teachers and leaders need confidence that the inspection system will treat leaders and teachers fairly.”

In a report, the company behind the artificial intelligence system says it’s hoping to work with Ofsted over the coming year to improve its algorithm and tailor its effectiveness to suit prioritising the underperforming schools that should be inspected.

The firm is using data gleaned from previous Ofsted inspections, census information and other school data; all of the information is already publicly available.

However, the algorithm also utilises the responses about a school from parents using Ofsted’s Parent View.

The report’s lead author, Michael Saunders, said that the data will not be shared with schools and he added it would not be helpful to do so.

‘Five things to help make your school better’

He told the BBC: “If we chased down the algorithm’s findings and offer five things to help make your school better, that will be disingenuous.

“Ofsted inspectors who do inspections are better placed to provide advice.”

Mr Saunders acknowledged that the algorithm could be used for future applications within education as well as a tool to help Ofsted.

He suggested that one issue could be the issue of GCSE predictive grades for a student as these are based on a teacher’s judgement but from research he says that these predictions are not accurate.

He added that the algorithm could give a more accurate picture to help youngsters in their education while any other applications would require ‘practical and ethical oversight’.

Help Ofsted inspectors prioritise which schools should be inspected

Mr Saunders says that the data will provide a low-cost tool that will help Ofsted inspectors prioritise which schools should be inspected with the algorithm highlighting which schools are likely to be failing but it cannot explain why.

He added that there will still be a need for human inspectors to find out what is going on and the algorithm has a high success rate to help identify schools that are deemed as being ‘inadequate’ or ‘requires improvement’.

The same use of data has also helped create an algorithm to identify inadequate GP clinics.