From bot exam markers to snooping surveillance tech, we look at some of the most eye-catching technology on sale to educators at this year’s ed tech Bett conference, as the regulatory world rushes to catch up with AI advances.
Bringing exam marking into the AI age
The makers of an AI tool which marks both handwritten and typed exam papers and provides detailed feedback, claim it will “revolutionise” how exams are marked.
Vision Marker is the brainchild of father-son duo Barry Lambert, a chief examiner who has written over 50 GCSE papers, and Dr James Lambert, a machine learning boff who previously ran an algorithmic hedge fund.
The pair are marketing their tool as a “co-pilot” for marking rather than a complete teacher-less solution. Dr Lambert claimed to be unaware of any regulations prohibiting the use of an AI tool as a sole exam marker.
However, Ofqual said this is “not allowed”, although “awarding organisations may decide that AI can support the marking process”.
The Lamberts claim their tool is capable of co-marking all 420 million questions submitted each summer for key stage two and GCSE maths and English papers.
Vision Marker is currently being used in 25 schools to co-mark such practice papers, and the Lamberts are working with exam boards as well as some multi academy trusts and higher education institutions around using their services in the future. However, they won’t reveal which ones.
AQA’s chief executive Colin Hughes said in November that his organisation was interested in the potential for AI to mark exams, describing it as being “very easy” to have “machines marking human markers”. But “public confidence” was stopping them from doing so.
Other AI-powered marking tools coming onto the market include Blees AI, based in Canada and London, Graide, a University of Birmingham spinout with a new platform for marking essays and reports for universities, and KEATH.ai, a platform developed by University of Surrey researchers.
But Dr Lambert claims Vision Marker is distinctive because it can mark handwritten exams, even noting where words are scribbled out, and in the quality of its feedback, which can come with “varying levels of encouragement”.
Dr Lambert believes this feedback could prove invaluable in showing college students retaking their maths and English GCSEs where they are going wrong. “The time pressures mean feedback is never given at exam level. But detailed feedback is essential for student development.”
While manual marking takes weeks to complete, which Dr Lambert claims “delays result to students and hinders their outcomes”, Vision Marker could deliver exam results “in a couple of hours out to the entire nation”.
He claims their tool has proven itself to be “more accurate than [human] markers”, who are “prone to error making in these highly repetitive tasks”.
Teachers use exam marking to top up their wages each summer, with AQA paying between £500 and £1,000 per exam. Some may be alarmed at the prospect of being replaced by AI. But Dr Lambert claims they are “paid below minimum wage for what they do” and that their tool “turns markers into moderators” as they “use the [learner] feedback to agree or disagree with our system.”
He believes this means their technology will not be subject to the “black box effect” currently concerning the central banking market, in which it can be hard to understand how a particular AI model arrived at its conclusion.
“It’s the regulator’s dream rather than their nightmare because we have a fully auditable process of how and why marks are awarded.”
Schools and colleges are using surveillance equipment to crack down on vaping, bullying and rowdiness in hidden areas, such as toilet blocks.
Companies are selling sensors to schools and colleges that can detect not just nicotine and THC, but also listen out for keywords and “noise incidents”.
Triton’s 3D sense pro and HALO Smart Sensors, made by the US company IPVideo (owned by Motorola Solutions), both actively listen for pre-programmed words through a machine learning algorithm. The sensor triggers an alert sent to staff members when they are heard.
The 3D Sense Pro sensor has 10 built-in keywords (such as ‘help me’ and ‘stop it’) it detects, and leaders can also choose ten “customisable” keywords it can listen out for. One company selling the 3D Sense Pro, Emergency Protection, described on its website how keywords and phrases were “constantly added through OTA [over-the-air] updates”. Similarly, Halo smart sensors come pre-loaded with five keyword phrases and educators can request to include others.
After the AI learns different pronunciations of new keywords, the software is updated so these words can be detected by all Halo sensors in schools and colleges nationwide, said Jon Glover, a manager at Halo sensor seller Millgate.
Kay Firth-Butterfield, a lawyer specialising in AI, said parents and young people should be contacted and asked for their consent first before this technology is rolled out.
But a sales representative for Schoolwatch, a UK seller of the 3D Sense Pro, said users did not need to get permission from parents to use the sensors, because they were not storing or collecting any personal information. The devices can integrate with CCTV too.
When activated by a vaping sensor, the cameras can capture every student leaving the bathroom.
Schoolwatch’s website describes the algorithm as “actively listening” to pupils.
The Halo 3C-PC version can even count the number of people in a room.
Madeleine Stone, senior advocacy officer for Big Brother Watch, said that “secretly monitoring bathrooms is a gross violation of privacy and would make students and parents deeply uncomfortable.
“No [education provider] should consider spying on private conversations and doing so is highly likely to be unlawful. This misguided surveillance poses a clear safeguarding risk and should be allowed nowhere near UK schools.”
FE Week put the spying concern to Schoolwatch’s managing director Andrew Jenkins. He said they had had “many conversations” at Bett about this issue, with “the main worry” being whether verbal exchanges were live monitored or recorded.
“The answer is no … when triggered, staff will receive an alert via SMS, push notification, and email. Nothing is saved, nothing is recorded, and nothing can be reviewed.”
A Triton spokesperson said: “The aim is to provide an additional layer of security against threats like bullying or sexual assault in these areas, reinforcing a safe environment. It’s important to communicate to students and educators that the system is designed to enhance safety, not to monitor everyday conversations.”
The 3D sense pro model, which includes the keyword and noise detector functions, are the most expensive (£999) and least popular sensor model that Schoolwatch sells. It has sold nine of these to schools, which Jenkins said are “focussed on stopping vaping not audio detection” while Millgate has sold 30 to 40 Halo Sensors to schools and colleges.
The number of young people using vapes has tripled in the last three years, with concern growing over the use of vapes laced with THC. Last year, East Surrey College was handed a ‘requires improvement’ rating over “poor behaviour” which included students vaping in the building.
A student at one college last year wrote on an online forum, the Student Room, how his college toilet vaping had set off a fire alarm and caused everyone to be evacuated. They said that on their level two construction course, vaping had “basically become the cultural norm”.
Meanwhile, on social media, students describe ways to outsmart the sensors by blowing smoke away from them and covering them in plastic wrap.
But some sensors use AI which recognises when it is being compromised and acts on it in hostile ways. IPVideo’s website says Halo sensors’ speaker and light can be programmed to “blink or shriek” when it is being tampered with.
A representative for Millgate said because the “AI technology” gets “used to its surroundings”, the sensor would send an alert if a student “tried to spray deodorant on it” to “mask” their vaping.
Halo sensors can send alerts when noise gets above a specific volume, and detect gunshots and explosions.
IPVideo’s website suggests educators can install them in locker room, classrooms and dorm rooms as well as bathrooms.
St Joseph’s College in Stoke-on-Trent placed Halo Sensors in two toilet blocks in September. After deputy headteacher Charlotte Slattery said they became “very active in the first few weeks”, it installed five others elsewhere. The school were “more interested” in the vaping alert than the sensor’s keyword and noise detection features, although they use those too.
A recent report in the LA Times found that in the US, the sensors are not always effective.
Michael Allman, a board member at California’s San Dieguito Union High School District, said that “in a way”, a pilot programme had been “too successful as the sensors went off so frequently, administrators felt it was useless to review security footage each time”.
At the Coppell Independent School District in Texas, sensors are part of a prevention strategy in which students can receive $50 for reporting vaping by peers. “They were turning each other in right and left,” said Jennifer Villines, the district’s director of student and staff services.
AI robot tutor
The learning platform Tassomai claims it’s new virtual tutor, Mai, can do almost anything that a human tutor does, but for a fraction of the cost.
Its founder, Murray Morrison, believes that one day soon, AI like his will do away with the need for tutors altogether, while levelling the education playing field in the process.
Mai uses a trained version of OpenAI’s GPT3.5 to chat to learners, in the same way that an on-demand human tutor would guide them with their homework.
Since January, Mai has been made available to all the 600 secondary schools in England that use Tassomai’s assessment platform.
Other companies in the formative assessment space include Oxford-based Educake which provides homework and revision, and the Canadian software company Showbie’s classroom assessment tool, Seneca.
All Mai’s conversations are recorded and stored by Tassomai.
Morrison admits there have been “lots of hiccups” in teaching Mai how to tutor, as large language models are “not as clever as you want them to be”.
“The art comes in training it to do the right thing…it’s certainly not just a plug in and play process.”
Last month, the delivery firm DPD had to disable part of its AI online chatbot after a customer was able to make it swear and write a haiku about how useless the company is.
Mai has been programmed to keep conversations “relevant and appropriate” to students’ Tassomai quizzes, using a moderation system whereby spamming, persistent inappropriate comments or questions from users are flagged. Those users have their access limited or blocked.
When asked by a learner recently if Mai had any “cheat codes”, the bot responded: “Haha. I wish I had cheat codes for learning, but unfortunately, there aren’t any shortcuts when it comes to gaining knowledge.”
And there are other ethical quandaries that have had to be considered. One school with a conservative Christian ethos insisted upon Mai not being able to teach some aspects of biology.
Morrison describes such instances as causing a “real dilemma” for companies like his.
“What if we were asked to write content for the Florida School Board, and they don’t want us to say that evolution is true? That’s a whole other conversation.”
There is also the issue of “half-truths or untruths” in the current GCSE curriculum which Mai is trained to guide learners on.
Morrison sees what is being taught by one exam board about “human-caused climate change” as “hedging its bets”.
Morrison believes AI is rapidly becoming more sophisticated. It is “not beyond the reach of AI” to be able to replace tutors altogether within the next two years. “When tutoring is as expensive and inconvenient as it is, there’s an opening there.”