DOES your favourite app seem to be taking longer to load than it used to? That may be due to ever-richer graphics and overloaded cellphone networks, which take their toll on smartphone apps and increase the time they take to boot and retrieve information from the network on, say, train times or the weather.

A way to make them boot faster, developed at the University of Massachusetts at Amherst, could mean your app might one day be ready and waiting for you the moment before you need it.

Tingxin Yan and colleagues have borrowed a trick from computer science to achieve speedier loading. Called predictive caching, it involves guessing which software routines are most likely to be needed for the next stage of a computerised process - so that the right app is primed to run when called on, without booting from scratch. The system uses the phone's location and motion sensors to learn when the user typically runs the app, Yan will tell the MobiSys conference in Windermere, UK, which starts on 25 June.

Imagine that, as you walk to a railway station each day, you normally get to a certain street corner and open a train times app to see if the trains are running to schedule. The software checks the time you usually do this, senses that you are walking and preloads the app, with the current train info retrieved by the time you arrive at the corner on which you normally request it.

In tests, the software cut 6 seconds from the average 20-second boot-up time for apps on Windows phones - although it gobbled 2 per cent of the battery per day while doing so.

App author Peter Bentley at University College London sees issues. "This works great if the things being cached are entirely predictable. But what happens in an emergency when you need to use a rarely used but essential app and the phone has preloaded lots of massive apps which then have to be cleared first?"

It would be better, he says, if coders designed apps to boot faster in the first place. It also depends on the type of app. "I would see this technology better suited to those apps that are designed to be used for 30 seconds or less and which need to go online and download/upload new data."

 An automated system that detects when online pupils are distracted or snoozing and then uses tricks to keep them alert

WE ALL remember dozing off during a boring class at school. A robotic teacher that monitors students' attention levels and mimics the techniques human teachers use to hold their pupils' attention promises to end the snoozing, especially for students who have their lessons online. Tests indicate the robot can boost how much students remember from their lessons.

Intelligent tutoring systems that use virtual teachers to interact with students could play a crucial role in the expanding field of online education. The trouble with online courses is that it is usually impossible to know whether the student is concentrating and engaging with the lesson. Unlike virtual teachers, human teachers have a series of tricks for keeping their classes focused - changing the pitch or tone of their voice, for example, or gesturing to emphasise points and engage with their audience. Bilge Mutlu and Dan Szafir at the University of Wisconsin-Madison wanted to find out whether a robot could use some of the same techniques to improve how much a student retains.

"We wanted to look at how learning happens in the real world," says Mutlu. "What do human teachers do and how can we draw on that to build an educational robot that achieves something similar?"

The pair programmed a Wakamaru humanoid robot to tell students a story in a one-on-one situation and then tested them afterwards to see how much they had remembered. Engagement levels were monitored using a $200 EEG sensor to monitor the FP1 area of the brain, which manages learning and concentration. When a significant decrease in certain brain signals indicated that the student's attention level had fallen, the system sent a signal to the robot to trigger a cue. "We can't do it just at any given moment, we have to try and do it like human teachers do," says Mutlu.

The robot teacher first told a short story about the animals that make up the Chinese zodiac, in order to get a baseline EEG reading. Next, the robot told a longer 10-minute story based on a little-known Japanese folk tale called My Lord Bag of Rice, which the student was unlikely to have heard before.

During this story the robot raised its voice or used arm gestures to regain the student's attention if the EEG levels dipped. These included pointing at itself or towards the listener - or using its arms to indicate a high mountain, for example. Two other groups were tested but the robot either gave no cues, or sprinkled them randomly throughout the storytelling. Afterwards, the students were asked a few questions about the Chinese zodiac to distract them before being asked a series of questions about the folk tale.

As the team had expected, the students who were given a cue by the robot when their attention was waning were much better at recalling the story than the other two groups, answering an average of 9 out of 14 questions correctly, as compared with just 6.3 when the robot gave no cues at all. The results were presented at the Conference on Human Factors in Computing Systems in Austin, Texas, earlier this month.

The idea of recapturing students' waning attention in this way would have "significant implications for the field of education", says Andrew Ng, director of Stanford University's Artificial Intelligence Lab in California and co-founder of online classroom Coursera. It offers free courses from Stanford, Princeton University, the University of Michigan and the University of Pennsylvania, and has already attracted more than a million students since its launch last month.

"One-on-one tutoring has been repeatedly shown to give dramatic results in student learning, but the main problem with it is the cost, and that it's just difficult to scale," Ng says. "The vision of automatically measuring student engagement so as to build a more interactive teacher is very exciting."