---

Work by the British anthropologist, Robin Dunbar and others suggests that there is a cognitive limit to the number of people each of us can form a relationship with. It’s usually quoted as approximately 150, and primarily dependent on the size of the human neocortex [1]. This number can vary because of limitations imposed by availability of food and other resources, of course, but the size of tribes, villages, and even units in the military have been attributed to neocortical limitations.

Given my experience with smart devices — from my smart watch to my voice-enabled smartphone to my smart home — I’m convinced that there is a Dunbar number equivalent for smart devices. Let’s call it the Maximum Intelligent Interface Count, or MaxIIC for now. Based on my presumably normal cognition, MaxIIC is between 50 and 100, depending on the modality and context of the interface with the smart devices.

Why does a MaxIIC matter? Well, given the rate of evolution and pervasiveness of the Internet of Things, AI, edge computing, robotics, and wearable computing, we’re going to be up to our eyeballs in smart devices — including social and service robots — within the decade. Moreover, with an aging population — including those whom social and service robot R&D is targeting — an increasing cognitive load is the last thing developers should be embracing.

There’s something inherently different about an intelligent interface to a smart device compared with a traditional interface. For example, compare that with the “dumb” interface to my tube-type stereo system. There isn’t a single processor in the component system, which I turn on and off by a foot switch located immediately next to the power amplifier. Similarly, I direct audio to different rooms in my house with a simple push-button distribution switch. The cognitive load of walking up to a device and stomping on the power button is minimal. I can envision having hundreds or thousands of such control configurations, the maximum number limited by the time it would take to walk from one device to another, not the volume of my grey matter.

Like many SERVO readers, I also have a smart home that includes voice controlled lights, appliances, computers, cell phone, and security system. Siri lives on my iPhone as one intelligent interface and the smart devices in my home, some of which are controlled by or at least accessed by Amazon’s Alexa voice recognition system.

When I retire at night, I have to remember which lights to turn off in the kitchen, how to set the alarm, and, for the bedroom, how to set the color and intensity of particular lights. For example, “Alexa, headboard 40%” turns on the lamp mounted on my headboard at 40% intensity. Other lights in my bedroom are simply on-off.

Similarly, I have five active desks: one for computer based work; one for freehand writing; one for ham radio; and two for electronics projects. I have to remember to call out “Alexa, turn on desk” or “Alexa, turn off workroom,” etc., to power-up the appropriate desk.

In this pursuit of automation, I’ve been forced to develop a relationship with the various fixtures in my house. And I’m beginning to feel the cognitive load — especially when I’m tired.

The MacIIC is clearly a function of the modality used to connect with smart devices. I’ve found speech has the greatest cognitive load because I must recall naming and function before barking out orders. Pointing and gesturing — very similar to the stomp switch on my stereo — has less of a cognitive load. However, this modality isn’t well represented in my house or in the industry. At least for now, voice interfaces are the most popular.

I’m not saying we should become luddites and avoid anything with a transistor or chip. In fact, I look forward to the time when computers and smart devices surpass human intelligence. However, with an aging population and proliferation of smart devices, we have a responsibility to insure that the number of robots and other smart devices we develop doesn’t overwhelm our cognitive capabilities.  SV

---

[1] https://en.wikipedia.org/wiki/Dunbar%27s_number

---