Jan Noyes works in the Department of Experimental Psychology because she is interested in how people interact with technology – their expectations and perceptions of it – particularly computers and the more advanced technology that might be found on the flight deck of a civil aircraft or control panels of a submarine. Much of her work is with engineers, because if you are going to design something that complex you need a multidisciplinary team. “I design from the perspective of the human users, whereas engineers design inevitably from the point of view of the engineering parameters,” she explains. “For example, automatic speech recognition works much better for men than women, possibly because it is designed by male engineers and they tested it on themselves.”
Amused, I explain about the little scenario I had experienced on the way in to see her. “What you have to do in that situation is look at the way we speak to each other: what happens when I don’t understand what you say? You speak louder and slower, you over-pronounce,” Noyes explains. “So when we talk to a machine we bring those expectations along with us.” Embarrassed, I recall that is precisely what I did. “But if you look at the way speech recognition systems work, they don’t operate on those principles at all. The algorithms are not working on the parameters that humans work on – so logically, it could be better not to have any expectations at all and treat the machine as an idiot.”
I design from the perspective of the human users
While it is difficult to get a machine to behave intuitively as a human would, Noyes finds that there are lots of little things she can do to improve software by carrying out experimental work with humans first. A big problem for speech recognition in the car, for example, is background noise – the engine noise, the radio, and noise outside the car all contribute to confuse an automatic system. So you can make the software more intelligent by studying how and in what situations a human actually wants to use it.
Another area Noyes has been particularly involved with is in the management of visual display systems on board civil aircraft. In these times of concern about the contribution aircraft are making towards climate change, there exists a strong rationale for an energy management system that not only helps the crew use the aircraft’s energy more efficiently, but also contributes to improved safety. Heavy landings, rushed approaches and runway over-runs could all be reduced through the provision of an energy display which allows the crew’s actions to be based on more accurate information. This in turn should lead to more appropriate and timely responses. For example, pilots lower the under-carriage ready for landing much earlier than they often need to, because airlines have to err on the side of caution. However, lowering the wheels causes tremendous drag, which increases both fuel consumption and noise pollution for those under the flight path. Improved information on a plane’s energy state during a descent could help the pilot decide on the best time to lower the wheels, without compromising safety.
Given the large number of factors governing an aircraft’s use of energy, this type of monitoring is best performed by sensor and computer technologies. So the problem facing Noyes was to find the best way of presenting all this information within the already busy cockpit. Since the concept of energy guidance was relatively new, there were no designs available on which to base a display, and although that was quite daunting, given the number of possibilities, the project had the advantage of being able to apply human factors from the outset. In collaboration with Smiths Aerospace, Cranfield and British Airways, three designs were evaluated. It was found that those with a predictive capability produced the most accurate decisions concerning aircraft energy states.
The idea is that you come in on a ‘good’ energy line
A predictive display is simply one in which the information provided is extrapolated forward in time, enabling the operator to anticipate what is likely to happen in a few minutes’ time. Perhaps surprisingly, predictive systems have been little researched in the past with only NASA reporting a sustained interest in the idea. The role of such a predictive display might be seen as being similar to that of the erstwhile flight engineer who, often with the aid of simple calculations, would make predictions about the state of the aircraft which helped the crew make decisions concerning the flight.The primary aim of an energy display would be to provide the pilot with an indication of the energy range in which the aircraft is operating and to ensure that current energy levels lie within an ideal energy range. The idea is that you come in on a ‘good’ energy line and if you deviate from that you are in what is called a ‘dirty’ state – a too high or too low energy state – the latter having more serious implications for safety.
Thus an important feature of the system is giving the pilot an indication of whether the deviation from ‘good’ can be corrected in time. So how to incorporate all this information into one system, while not overloading the pilot? Interestingly, experiments showed that while a high workload only adds to the general ‘noise’ of the flight deck, a low workload display appears to be highly erroneous in terms of decisions made based upon it. Either way, potential for error is high, thus a balance between accuracy of decisions and workload is needed. Of the three systems evaluated, the ‘total energy status and predictive display’ – or Energy Line – shown above appears so far to be the most effective, but the aircraft industry does not move fast and designs of today may take many years to bear fruit.
I asked Noyes if she thought there would ever come a time when humans became redundant in aircraft. “My understanding is that it is very feasible to have unmanned aircraft – the advantage of crew is when things go very wrong. Humans have an incentive to get home, whereas a computer doesn’t,” she replied. “The joke is that aircraft of the future will only have a man and dog at the controls. The man is there to feed the dog. So what’s the dog there for? To keep the man off the controls!”