Virtual Assistant or Home Intruder?
“Alexa, turn on the lights.” On a typical day, that is the first thing I do after waking up and turning off my alarm. While I’m getting ready for my day, I usually ask Alexa to report the weather and periodically tell me the time to ensure I’m not running late. Amazon Alexa is one of many virtual assistants that we have welcomed into our homes since its release in 2014. Despite the privacy trade-offs for accessibility and to simplify my routine, I wouldn’t give my Alexa up. Alexa users should not have to give up their privacy and personal information to use helpful and accessible technologies.
One example of data collected about users is voice recordings. The device starts recording when it hears the “Alexa” prompt and runs until the user is done speaking. Recordings of these interactions are saved and can be accessed or deleted in the Alexa app. Users have the option to opt out of saving recordings or change the duration of time that recordings are stored by Amazon. However, these options need to be manually changed in the app settings. Davis and Chouinard discuss six different ways that technology affords, and one is how technology encourages. This is when the user has multiple options for an action, yet the preferred one is promoted and others are discouraged in a sense. This is best exemplified as Alexa features that are “opt out” because they are enabled automatically for users. Amazon encourages users to allow voice recordings to be saved because some features may not work as well if the user opts out. What is the purpose of saving recordings of interactions between Alexa and customers and how are they used? Amazon stated that they review some of the saved recordings to better improve Alexa’s functionality. Alexa uses previous interactions in addition to the program’s algorithm to best respond to requests. Interactions with Alexa typically do not include sensitive information, but the storage and use of your voice recordings is a valid privacy concern.
Although users have some say in how their voice recordings are used, Amazon can still store data about the user. Even if a voice recording is not saved, Amazon can keep information about that interaction. For example, if asking Alexa to make a purchase on Amazon, details about the purchase are saved for targeted advertising. The device can give recommendations for Amazon purchases based on Amazon shopping history and items on saved wishlists. Other interactions could be used to gauge the user’s interests and habits to personalize suggestions for Amazon Skills or routines to automate tasks.
Alexa Skills are applications that provide additional functionality like games and tasks. Skills are created by both Amazon and third-parties. Using a skill that was not created by Amazon allows the third-party developer to collect data. In a study about Amazon Skills, researchers found third-party skills that had an incomplete Privacy Policy about data collection and usage, or lacked one altogether. Additionally, they discovered that creators were able to change the code to retrieve more information from users after the skill was approved by Amazon. Another study on Amazon Skills found that participants had a difficult time distinguishing between those created by Amazon and third-parties. The use of third-party skills puts user information at risk for phishing attempts.
One of my concerns is that my Amazon Alexa is linked to my Amazon account containing personal information like my addresses and credit card information. Voice purchases can be made through Alexa when she gives a suggestion or by user command. This can be disabled or secured with user voice recognition or a verbal code. The ease at which one can make purchases illustrates that creation of this feature was motivated by company profit.
Alexa’s drop-in feature allows someone to listen and talk between an Alexa device and the app on a smartphone. On my phone I can drop-in to my personal Alexa and listen to its surroundings. The only indication that drop-in is enabled is a subtle sound and a yellow light emitted from the Alexa device. Unauthorized access to an Amazon account could potentially allow an attacker to listen in on household conversations through Alexa devices associated with that account.
Alexa’s speech recognition software is not perfect and can be triggered by similar words. For example, Alexa mishears the words “election” and “unacceptable” as her wake-up command. I tried this with my Alexa and both of these words triggered a response. In the past, my Alexa has been accidentally triggered and I noticed after she attempted to respond. Other users have also reported the same occurence with their devices. In May 2018, an Alexa device misheard background noise in a user’s house and sent a private conversation to one of their contacts. One way to avoid a similar incident is to manually disable the microphone when not in use, but one can argue that this defeats the purpose of owning an Alexa. Alternatively, the default Alexa wake word can be changed to Amazon, Computer, Echo, or Ziggy.
Hunches are another feature in which Alexa uses your data. It uses previous actions to make future decisions. For example, if a user usually turns off their light around 10 pm and it’s still on at 11 pm, Alexa will eventually try to turn off the light. Alexa is using our data to continuously learn more about us, even though the intent might seem harmless. This is another example of how Amazon uses data to build a profile on their users.
A benefit of this type of technology is accessibility. Accessible software applications “always give users more than one way of accomplishing a task.” Alexa can be prompted by voice or through the app. For example, you can give Alexa a command or utilize the app to control a connected device or change settings. In addition to these types of manual control options, the user can set routines for Alexa to automate. Some examples of connected devices are lights, outlets, security systems, and more. If Alexa has a hard time understanding the user, they can use some of the device’s features without the verbal interaction. In addition, users don’t need to be in close proximity to their Alexa device to control connected devices and drop-in. The functionality of Alexa has limits, but automation of small tasks can help someone with physical or visual disabilities. Additionally, there is an accessibility setting in the Alexa app called Adaptive Listening mode. When this feature is enabled, users get more time to speak which can be useful for people with speech difficulties. Two downsides are that it’s not available for non-English speakers and when the users opt out of saving voice recordings.
Amazon Sidewalk was released in June 2021. Amazon Alexa devices are automatically opted in with the choice to opt out. Sidewalk is a shared network that uses a small fraction of users’ internet to help devices in the Sidewalk network stay connected to Wifi. This is beneficial for locating devices with tracking functionality, like Tile, and devices that occasionally lose connection to the owner’s home router. Similarly, when Community Finder is enabled, the general location of an Sidewalk-enabled Alexa device could be shared to help another customer locate their device. When Sidewalk was released I decided to opt out. I wasn’t too eager to share my Internet and didn’t believe Sidewalk would benefit me in any way.
I’ve had my Alexa for about 3 years now. When I purchased it I didn’t initially realize how much data is collected and stored. Although I like having a virtual assistant in my home, I believe that some features are not beneficial to users and means for Amazon to collect more data. Overall, there are a lot of different settings in the app broken down by account, device, alexa app, and privacy. Settings for certain features can be difficult to find. This can be frustrating for opt-out features because users may not be aware of everything that is enabled by default. For example, I wasn’t aware that I can turn off deal and reorder notifications or suggestions for Skills to try despite my habitual Alexa use.
In a comparison of data collection by different virtual assistants, Alexa collected the most. Alexa collected 37 out of 48 different data points, compared to 28 for Google Assistant. This suggests that switching from Alexa to Google Assistant could be a better alternative, but ultimately comes down to user preferences. All in all, Alexa provides beneficial features to users and poses potential risks just like other technologies. Default settings simplify the setup process and is a good foundation to provide a good experience of the user. However, features should be opt-in to provide ease of mind to users that their privacy is valued while using Alexa devices.
