Most of us – not all – figure out an approach to workplace behaviour while on the job and through trial and error.
Through the cut and thrust of everyday work we learn when it’s the right time to talk to colleagues and when it isn’t. We learn that some colleagues are happy to talk any time and some hardly at all, and that some don’t ever want work conversations to stray into personal territory.
It’s going to be increasingly important to “teach” such routine human skills to robotic systems as we move towards them filling a variety of human workplaces.
This isn’t simply a matter of programming them to understand commands and to ask simple questions (think Siri). They have to learn about two-way communication – about saying what they’re doing and hearing what others are doing so that everyone works together effectively.
With that in mind, researchers at Massachusetts Institute of Technology’s (MIT) Computer Science and Artificial Intelligence Laboratory (CSAIL) have developed a framework called CommPlan, which gives robots a few high-level principles for good etiquette and then leaves it to them to make decisions that allow them to complete tasks as efficiently as possible.
CommPlan, they say, contrasts to more explicit, handcrafted communication policies that would see robots interacting using hard-and-fast rules based on specific probabilities. This difference between the two approaches is subtle but important.
CommPlan uses learning and planning algorithms to do real-time cost-benefit analyses on its decisions.
For example: will asking a human colleague a question save time by ensuring the robot doesn’t do the wrong thing, or will it slow down a work process because it stops or slows the human from performing their task? The robot might weigh a combination of factors, such as whether the human is busy or likely to respond given past behaviour.
In contrast, a handcrafted policy essentially has to be designed anew for each task and context.
The MIT team tested three approaches – CommPlan, a handcrafted policy, and a communications-free silent policy – as part of a kitchen scenario involving tasks such as assembling ingredients, wrapping sandwiches and pouring juice.
Experiments, they say, showed that the human-robot teams performed more safely and efficiently using CommPlan compared to either of the other options.
Vaibhav Unhelkar is one of the lead authors on a paper about CommPlan that will be (virtually) presented in coming days at the Association of Computing Machinery / Institute of Electrical and Electronics Engineers International Conference on Human-Robot Interaction.
He says he’s encouraged by the success of CommPlan, since handcrafted policies require significant time, effort and expertise on the part of programmers.
“CommPlan combines the power of human experts and algorithms to create policies that are better and at the same time require reduced developer effort,” he says.
Additionally, the researchers say that a handcrafted policy’s reliance on cut-and-dried rules makes it more likely to suffer from hiccups such as overcommunication.
“Many of these handcrafted policies are kind of like having a co-worker who keeps bugging you on Slack, or a micromanaging boss who repeatedly asks you how much progress you’ve made,” says co-author Shen Li.
“If you’re a first responder in an emergency situation, excessive communication from a colleague might distract you from your primary task.”
CommPlan’s performance offers great promise for its applications in other domains, such as manufacturing and, especially, healthcare.
In a world feeling its way through the COVID-19 pandemic, in which healthcare workers are among the most vulnerable because of their need to interact with patients and one another – CommPlan could help speed the pace at which robots enter human workplaces to assist with tasks that would otherwise expose or endanger humans.
The CommPlan team has so far only used the framework for spoken language, but they say it could also be applied to visual gestures, augmented reality systems, and other approaches.