Human machine interfaces include things like the cockpit of an aircraft or a space ship, but also plain old things like the gauges and knobs for the radio in your car. These interfaces must be designed so that they’re easy to use, and leave enough attention to focus on for example the road.

cognitive load eye tracking cartoon

Eye tracking can be used to investigate the effectiveness of human machine interfaces. The most important thing to investigate is congitive load. When the cognitive load of a user is too low, the user will be bored, in a state of low arousal, and will be slow to react to dangerous stimuli in the environment. When the congnitive load is too high, users will start to experience visual tunneling. The attention of the user will narrow, and stimuli that would otherwise be easily noticed, go unnoticed. When congitive load is too high, users of the interface will also take more time to read out instruments. This is reflected in the ‘dwell time’, as measured with eye tracking tools.

Measures

 

There are a couple of measures that can be read out from an eye tracking experiment. Before starting the experiment, think about the tasks to give to the subjects you want to measure, so that these measures will tell you something about cognitive load in different situations.

 

  1. Blinks per minute (more blinking = higher congitive load). Easy to measure, but not super informative on its own.
  2. Dwell time (higher dwell time = higher congitive load). This is the most important measure and easiest to link to actionable conclusions.
  3. Time till focus. This is not always measurable, and can only be measured with stimuli that appear and disappear. It is good to add these stimuli to the experiment, so visual tunneling can be indexed. If stimuli appear but are missed, or time till focus is very long, congnitive load is too high.
  4. Focus priority. When stimuli compete for attention, eye tracking experiments can determine which stimuli will attract attention most readily, or which gauge is read out first by users under which experimental conditions.
  5. Relative total focus time. During the experiment, the first thing you will do is mark regions of interest. It is informative to see where the subjects in the experiment focus most on average over the total course of the experiment, or in certain experimental conditions. Maybe some gauges are not even used at all for example.
  6. Revisists. As opposed tot total focus time, what is the number of times the subject looked at this region of interest. Some gauges are more difficult to read out, so they will get a high focus time. Some gauges are very fast to read out, get a low total focus time, but might be used very often, very briefly.

Before you start: Scenarios

 

There is one thing that has to be determined in advance of the experiment that is very important to get right: The different scenarios.

These scenarios are marked during the eye tracking experiment. The experimenter presses a button in the companion app, or the start of the scenario is signalled with a Python script, JavaScript, Arduino script, or other script that used the OpenEye libraries to send events to the experiment.

This is important because these events cannot be added later so easily. Adding them in a thought out manner from the start of the experiment will make it easier to later process the eye tracking data fully automatically.

Regions of Interest

 

Also important are the regions of interest you define. These can be altered after the experiment however, so if you want to add more later, or remove some that turned out to not be so informative and cause clutter, this is no problem.

With OpenEye, regions of interest are added using artificial intelligence. This makes it possible to track all sorts of things. Here are some examples of things you can define as regions of interest:

 

  1. A specific gauge.
  2. The window, of a car for instance.
  3. A face / head, of a copilot for example.
  4. Warning lights, they will be detected when they turn on, important for ‘time till focus’ measures.
  5. The entire user panel, to measure for example the ratio of looking out the window and at the panel.
  6. The side mirror.
  7. Other cars on the road, yes, moving objects can be defined as regions of interest.
  8. And more …

 

Once regions of interest are defined, together with the start events of different scenarios, all kinds of measures pop out of the OpenEye analytics software. Some measures you can directly use from the OpenEye analytics software, other metrics that combine different regions of interests for example, can be easily created because all measures are also given in a format compatible with Excel and Sheets.

 

If you have any questions on how to do a proper and informative analyisis of human machine interfaces with OpenEye, we are always happy to help. We are always interested in all the different experiments OpenEye is used for.