{"id":92267,"date":"2016-05-14T12:00:27","date_gmt":"2016-05-14T10:00:27","guid":{"rendered":"http:\/\/mygaming.co.za\/news\/?p=92267"},"modified":"2016-05-14T08:40:11","modified_gmt":"2016-05-14T06:40:11","slug":"how-to-design-trustworthy-robots-that-wont-deceive-us","status":"publish","type":"post","link":"https:\/\/mygaming.co.za\/news\/gadgets\/92267-how-to-design-trustworthy-robots-that-wont-deceive-us","title":{"rendered":"How to design trustworthy robots that won’t deceive us"},"content":{"rendered":"

Does your car \u201cnot want\u201d to start on cold mornings? And does your toaster \u201clike\u201d burning your toast? This kind of intentional language is natural to us and built into the way we interact with the world \u2013 even with machines.<\/p>\n

This is because we have evolved to become extremely social animals, understanding others by forming mental models of what they are thinking. We use these skills to understand the behaviour of anything complicated we interact with, especially robots.<\/p>\n

It\u2019s almost like we believe that machines have minds of their own. And the fact that we perceive them as intelligent is partly why they have such potential. Robots are moving beyond industrial, commercial and scientific applications, and are already used in hospitals and care homes.<\/p>\n

Soon it will be normal to interact with robots in our daily lives, to complete useful tasks. Robots are also being used as companions, particularly for elderly patients with cognitive impairment<\/a> such as dementia. After years of scientific study, this has proven very successful<\/a> at improving long-term quality of life.<\/p>\n

However, there are ethical concerns about vulnerable people forming relationships with machines, in some cases even believing them to be animals or people. Are robot designers intending to deceive patients? As robots become more important to us, how can we trust them not to mislead us, indeed should we trust them at all?<\/p>\n

In 2010, a group of academics produced ethical guidelines for how we should build robots, much like science fiction writer Isaac Asimov\u2019s famous laws<\/a>. Asimov stated that robots could not do anything to harm a human being; that a robot should always obey a human; and that a robot should defend itself so long as this didn\u2019t interfere with the first two rules. Similarly, these academics produced the EPSRC Principles of Robotics<\/a>.<\/p>\n

For me, the most interesting principle is number four: \u201cRobots are manufactured artefacts. They should not be designed in a deceptive way to exploit vulnerable users; instead their machine nature should be transparent.\u201d<\/p>\n

Hidden decisions<\/h2>\n

In my research group, we are conducting experiments with robots to investigate how well we understand them. Ultimately, we want to find out how we can best design robots to improve our mental models. We\u2019re trying to show that it\u2019s possible to create transparent robots that are useful, and when necessary emotionally engaging, despite having a transparent machine nature.<\/p>\n

We assert that if we can make robots more transparent, then we won\u2019t need to trust them, we\u2019ll always know what they are doing.<\/p>\n

\"Little<\/a><\/p>\n

We use a simple robot that moves around a room avoiding objects while searching for humans. When it finds a human it flashes lights, does a small wiggle dance, and then trundles off seeking another one. Sometimes when in a corner it goes to sleep to save battery. That\u2019s all it does.<\/p>\n

We videoed the robot, showed this to a group of 22 people and asked them what the robot was doing and why.<\/p>\n

Some of the answers were remarkable. Based on cues from the environment, and the imaginations of the people, they came up with all sorts of ideas about what the robot was up to \u2013 views that were generally quite wrong.<\/p>\n

For instance, there is a bucket in the room, and several people were sure the robot was trying to throw something into it. Others noticed an abstract picture in the room and wondered if the robot was going to complete the picture.<\/p>\n

These people were mainly graduates in professional jobs, and several had science, technology, engineering or maths degrees. Almost all used computers every day. Although we did not program the robot, nor create the room explicitly to mislead, the observers were deceived.<\/p>\n

\"Robot<\/a><\/p>\n

We showed the same video to an almost identical second group. However this group was simultaneously shown a display demonstrating each decision in the robot\u2019s action selection system controlling its behaviour, synchronised with the robot moving around the room and tracking objects.<\/p>\n

It is a kind of dynamic heat map of the processes and decisions being made inside the robot\u2019s brain, making plain the focus of attention of the robot, and outlining the steps it takes to achieve its goals.<\/p>\n