The horseless doorbell

The Internet of things should spell the end of doorbells as we know them. There, I’ve said it, and I’ve said it up front, because hearing about what a complete fail my own IoT doorbell has been is not likely to lead you to that conclusion on your own. That story now.

You will recall I set out to build an IoT doorbell an embarrassingly long time ago. That project produced some interesting musings but not a reliably functioning door-adjacent alerting device. Building a prototype was easy enough, indeed, a matter of minutes with LittleBits components I had at hand. What I was unable to do was get from prototype to a reliable, always-on version. Here’s my final, throw-in-the-towel model:

Note the ventilation piece over the network chip. Details, details!

This picture well illustrates the three issues that plagued me, only one of which I eventually overcame. Those issues were, in order of difficulty, mechanical, human, and network:

• Things fall apart: The main selling point of LittleBits is that they snap together magnetically. But that which magnetism has connected, any man, or, more likely, boy, can all too easily disconnect. In fact, the small mechanical force required to push the button was enough to interrupt the flow of electricity to the network card (green status light, above) responsible for reporting that push. Designing a case capable of holding everything rigidly in place was really surprisingly difficult. In fact, had I not had access to infinite Lego and a 9 yo master builder, I doubt I would have solved even this problem.
• Humans don’t see what they don’t expect to see: My household orders its fair share of goods online (and I order what some would say is more than my fair share of vinyl LPs that way, too), and during the period my prototype was up I paid close attention to deliveries to see if they correlated with doorbell rings (and to ensure my LPs didn’t melt in the sun). They did not. Ever. Nor did the visits of anyone who hadn’t already had the doorbell pointed out to them. In other words, people didn’t see the bell, despite my perfectly legible gray “Bell” label (see photo, above, in case you didn’t notice it already, which you did, right?).
• The network is opaque: Pinning down all the parts ensured I always had a status light. Periodically resetting my home network meant I often enough had a green (as opposed to red or, more often, purple) status light. But what to do when I had a green light but still the button push did not register on phone, email, or Twitter as, at various points, it should have? This happened often and was utterly mysterious when it did: being a normal human being, I don’t know where to find the logs for my home network, and not being an employee of LittleBits, IFTTT, or Comcast I don’t have access to those logs, either. When the system didn’t work, I didn’t and couldn’t reasonably know why.

From this I draw these lessons:

• The technical and organizational opacity of a heterogeneous IoT network is a major block to debugging it. This inhibits development and adoption of consumer IoT devices generally. To some extent this must also account for the motivation to offer, and to adopt, One Household System To Rule Them All packages.
• A buggy IoT device produces incomplete data, which seriously undermines the value of dependent analytics (that having been said, this experiment did alleviate my concerns that people were often ringing my original low-volume mechanical doorbell but going away, unnoticed). We cannot come to rely on our things to take care of us — to let the right people in and keep the wrong people out, for example— without a very high degree of trust in them. Less obvious, a traitorous doorbell undermines our trust in all things.
• In order to take advantage of existing habits (or, more pointedly, to avoid having to make people changes those habits), IoT devices offered as a one-to-one replacement for dumb but long satisfactory devices should not adopt novel forms.

So why, given a failed experiment, do I conclude that what the IoT really means for doorbells is their end? Two reasons. First, all of these problems are solvable, albeit not by me, and it’s the solving of just these sorts of problems that produces a paradigm shift in technology. Second, my experiment was flawed from the get go: we shouldn’t be seeking to replace dumb devices with smart ones in a like-for-like swap. This represents the “horseless carriage” approach to car making, i.e., the initial inability to think beyond a long-established mental framework when innovating around a new technology.

The classic doorbell is a soon-to-be-outmoded response to the need to be notified of the presence of a visitor. It is a suboptimal solution in at least this one respect: it requires the voluntary participation of the visitor. Turns out the UPS delivery person does not often choose to participate, nor, I suppose, does the thief. My mailbox, placed in accordance with US Postal Service regulation at the perimeter of my property, if made smart and given the right senses, could report on those crossing that perimeter. Given more advanced capabilities, those reports could contain enough data to support analytics that determined which of those events — UPS guy, yes; visiting cat, no — were worth “ringing” me about.

No, patient reader, I am not now setting off to build an IoT mailbox (though I welcome reports of same). My project card is full at the moment, and another topic, even more intriguing, has come to my attention. Please stay tuned for an as-time-permits report on “IoT Inside: The DIY Pancreas.”