Teaching machines to think instinctively

The last year or so has seen considerable discussion given to the interface between man and machine.  Undoubtedly the principle area for these discussions has been around autonomous vehicles, with discussions examining how they will interact with human road users, and of course how human passengers may regain control of their vehicles.

Autonomous vehicles aren’t the only domain that is seeing increasing levels of automation operating alongside humans however, and a recent paper examines the issue in more depth.

Thinking fast and slow

The paper revolves around the conscious and unconscious ways of making decisions that were so famously highlighted by Daniel Kahneman.  The paper argues that most attempts to understand the interaction between humans and machines focus unduly on conscious, deliberate decision making and neglect our more intuitive side.

As Kahneman reminds us though, most of our thinking tends to be of the faster, instinctive kind, and the paper argues that automated systems will be significantly improved if they incorporate intuitive as well as analytical cognition.  To support this, a new dual-processing taxonomy is proposed based on the work of Raja Parasuraman.

Intuitive cognition,” the authors says, “should be encouraged whenever automation fosters a quick grasp of the meaningful gist of information based on experience or perceptual cues, without working memory or precise analysis.”

For instance, if a human is working with a computer, it might better engage our instinctive side if the interface is in graphical form rather than textual form.  Not only is such thinking more common than its rational peer, it is also often more effective.

“Intuitive cognition is relatively immune to time pressure and workload, unlike analytical cognition, which is slow in responding,” the author says.

Of course, ensuring this level of cognition is present in automated systems is something that has long baffled computer scientists, but the paper believes that the human-automation taxonomy he proposes that encompasses instinctive cognition can play a useful part in the process.


Leave a Reply

Your email address will not be published. Required fields are marked *