Are We Paranoid about Digital Assistants in the Home?

on May 15, 2019
Reading Time: 5 minutes

Over the past few weeks, there have been different stories around digital assistants and eavesdropping, and about Alexa in particular. The microphones and cameras that are proliferating in our homes have been a concern since the inception of the connected home. Our concern with cameras and microphones does not start there though. People have been paranoid about those for years and went through the practice of putting tape on their camera and silencing their microphones. So much so that in 2019, the camera’s physical shutter has become an applauded feature of many enterprise notebooks.

When it comes to the home, our level of concern grows. Trying to think logically as to why that is the case might not be the best way to find an answer.

It Is not about Technology

I think that Amazon has been quite mindful, from the very beginning, about the level of trust that putting a device like an Echo in the home requires. Alexa’s blue lights were undoubtedly designed to increase our comfort level by signaling two quite subtle things: Alexa was hearing us, and Alexa was listening to us. Although it seems the same, these two are two separate things. Alexa must hear your voice, and Alexa is listening to what you say to act on your command after she hears the wakeword.

What I think is less clear to most consumers, is how Alexa and the cloud communicate and the link between what you say, what Alexa hears and what Alexa then hands over to the cloud to be processed. To get an answer, people continue to believe that Alexa listens, all the time. And therefore, Amazon does too!

This renewed attention to what digital assistants can hear in our homes was first driven by news that a group of Amazon employees and contractors are tasked with listening to Alexa’s recordings to transcribe them, annotate them and then feed them into the model to make Alexa smarter. Amazon responded to the article by explaining that only a small number of recordings is annotated and that employees do not have direct access to information that can identify the user or their account. All information is treated with high confidentiality with the use of multi-factor authentication and encryption. Amazon also gives users the option not to have their recording used to improve Alexa.

Using humans is not a practice limited to Amazon, both Apple and Google use similar methods. Apple reviews recording without personally identifiable information and stores them for six months tied to a random identifier. Google accesses some audio from its assistant, but it’s not associated with any personally identifiable information, and the audio is distorted, and the data is randomized.

Last week Geoffrey Fowler wrote an article for the Washington Post  discussing the finding of his recordings investigation. Every user can access those recordings through the Alexa app under the privacy settings, so I went to take a look at mine. I found that there were three categories of recordings:

  • Clear commands: Alexa set a timer, Alexa stop, Alexa play….
  • Unknown: where I usually found TV recordings that were mostly intellegibile
  • Audio was not intended for Alexa. This was by far the category that was most interesting as it involved random conversations around the house.

This last category is the one that I am sure most readers who use Echo devices would be creeped out about. Yet, in the review of my entire history, I did not find any snippets to be longer than a few seconds, certainly not enough to be meaningful. The most exciting revelation of this exercise was how many timers a human being can actually set thanks to Alexa with the implicit question of how this task was performed before we had Alexa!

It Might Be Meaningless Info, but It Is Still Info

In my case, the information was all meaningless, but to be honest, I do not think it is the value of the information that in this case should determine who gets to listen to it or gets to use it. It is fascinating, however, that in a study we, at Creative Strategies, ran in 2017 only 5% of the 800 American respondents said that privacy mattered in connected home devices. This compared to 60% who said smartphones and 30% who said PCs. If you think about it logically, on average, there is much more sensitive information on a smartphone or PC which fully explains our results.

So why the concern now? Because it is our home and we think our home is private by default, we even say “in the privacy of my own home.” So feeling like this is no longer the case because of technology is a compromise too many for some consumers. I think the same can be said about the concerns around the human component. Remember when you had to make calls using an operator who could listen to all your conversations? I don’t cause I was not alive then, but plenty of old movies have scenes like that. And what about every time you take a Lyft or an Uber, and they know your name and address. Or again, every time you have an inquiry with a service provider or a government official, and you have to share your social security number — all these examples of exchanges of relevant information. The difference here is that all those exchanges are informed.

Transparency and Controls

At the end of the day, as it is the case with all the data we generate, being social media, smart home, smart office, smartphones, PCs, we want to be aware information is being collected. On top of that, we want to be able to decide if we are happy with it or not and we also want to be able to change our mind about it. And finally, we want to make sure our data is secure and not misused.

Most brands give a good level of control to users. You can decide not to store your recordings and not to help improve their assistant. I also believe that as smart speaker, and smart homes, penetration grows and moves beyond early adopters a more precise explanation of what happens when you initiate a request with Alexa, the most used assistant in the home, would be beneficial. I know some people might want to suggest Alexa could add more warnings, so people are more aware of her presence. But it is a delicate balance Alexa needs to strike between making herself known and becoming a nuisance. I do wonder if Alexa having an “incognito” mode like what Google announced last week for search and maps could help her case.

I am sure we will see more experimentation in this area. We need to remember that for Amazon, a higher level of trust means higher engagement with Alexa which in turns drives more revenue. So even if you are more skeptical than I am about Amazon’s intent, I think you would agree with me that it does not make business sense not to strive for a high-level of trust.

Lastly, I also cannot help but think that what happened with Facebook has impacted consumers’ trust across the board and has put other brands under the microscope. The reality is that Amazon, Google, Apple and every other brand that “sells” you a smart device or a smart solution will need data about you to create such a thing. Data is what powers AI, so you first must decide if your data is a currency you are willing to spend to live a smart life and then you must decide who is worthy of it.