Smartwatches, smart homes, smart robotics… Today, computer technology is incredibly smart. But how do we ensure that we - as users - understand at all times what exactly these super-intelligent systems do, why they do it that way and how they work. And how can we ensure that we remain in control? In addition to intelligence, we also need intelligibility.
Many problems in Human-Computer Interaction have arisen with the shift from conventional desktop PCs - which do not perform any action without us commanding them via the keyboard or mouse - to smart environments that initiate actions autonomously. Home automation, smartphones and Fitbit bracelets are examples of this. These days we often don’t even spontaneously recognise the carefully hidden computers that surround us as computers. They continuously collect data via sensors and make independent decisions based on that data. They often make our lives easier with what they do, but at the same time, these systems are also becoming a lot more difficult to read and control.
Exactly what data do these hidden computers collect? What do they do with that data? On what basis do they make a decision? And how can we ourselves maintain control and adjust the system if necessary? That is essentially what intelligibility is all about. To make this generation of intelligent, complex, networked computing systems more usable, it’s no longer enough to simply improve the visible user interface. In addition, we must also make the hidden processes visible and ensure that these systems explain themselves better. Only when we succeed in this will we have created systems that are not only intelligent but also intelligible.
Sometimes intelligibility impacts privacy. An example of this is the bluetooth weighing scale. Many people who bought these scales did not realise that the device was connected to the Internet via their computer or smartphone, and that the scale was sending data to a cloud service that could then sell it to marketers. When consumers suddenly received advertisements in their mailbox for weight-loss products, they felt they had been duped. And rightly so. The system should have made it clear from the start that it would collect data and share it with vendors of weight-loss products. In addition, the system should also have given consumers the choice not to share their data. Opt-out should become a basic right.
In addition to privacy, intelligibility is also about providing systems that are more usable. There is still a lot to be achieved in this area. At the moment, technology companies usually opt to keep the user interface as simple as possible, preferring to keep the underlying processes and complex algorithms neatly concealed behind it. The rationale is that the average user would find them bewildering. There is some truth in this, of course: if, for example, Google Translate were to show you every underlying algorithm, as a non-computer scientist you would be completely swamped with all that information. And yet the system would work even better if it made a number of aspects visible to the user. Why has it chosen to translate a word exactly that way at a certain point? In what other contextualised sentences has it searched? If you make that information search partially visible, users can often find an even better translation themselves.
The application possibilities for intelligibility are virtually unlimited. It can even make a difference in the manufacturing industry. Robots are increasingly being used in the workplace. Man and machine literally work side by side on the same production line. In such circumstances it is of course very important for these employees to know what their robot colleagues are going to do, so that they can anticipate their behaviour. Intelligible systems can have a major impact in this kind of production environment in terms of safety and efficiency. A lot of attention is already being paid to safety. For example, most robot arms slow down when people come too close, but it is disappointing that the specific zones that the robot moves to are not shown to the production workers in advance. If you make these zones visible, employees will be much more careful there, which can only be a positive thing for efficiency and safety.
Internet of Things applications are quickly becoming commonplace, allowing users to automate all kinds of tasks themselves. It is relatively easy to connect your doorbell to your intelligent lamp so that it changes colour when someone rings the bell. But in a world where computer systems are increasingly interconnected behind the scenes and data is constantly being exchanged, intelligibility is becoming more and more important. How do we ensure that users can maintain an overview and stay in control at all times? How can we also add intelligibility to all these intelligent systems - even if they were not originally designed for it? We will continue to research this in the years to come.