S. A. Applin, Ph.D. is an anthropologist whose research explores the domains of human agency, algorithms, AI, and automation in the context of social systems and sociability. You can find more @anthropunk.
At I/O Tuesday, Google rolled out new features for Google Assistant. Google claims that these new features will be “helping you get things done to give you time back.” The announcement listed many improvements and options to streamline life with Google software and hardware. For starters, Google now thinks that teaching kids to ask for a movie by saying “Pretty Please” will ingrain politesse rather than teach children to beg from an Autonomous Agent. That said, I’m going to skip that for now, and focus on what Google is positioning as its future of communication with Google Assistant.
The company seems to feel confident that Google Assistant is now ready to help users save time with tasks such as buying movie tickets or ordering a coffee. This has been expanded with Google Duplex, billed as “an AI system for accomplishing real world tasks over the phone.” The Duplex technology underlying Assistant is now claimed to understand “complex sentences, fast speech, and long remarks.”
The goal of Duplex seems to be targeted towards making calls to make appointments with the people answering the phone in restaurants, hair salons, and other small local businesses that currently employ a human to answer the phone.
Google claims that this will be helpful to small businesses, which “rely on phone calls for appointment booking,” the company says. The goal is to give people (e.g. Google customers who do things mostly online and who don’t like to make calls) a mechanism by which they can interface with businesses that do not currently offer ways to mesh with the Google customer’s preferred communication habits.
In one interpretation of this, small businesses would gain more customers, with Google’s technology handling the details so that people who manage their tasks asynchronously and who have embraced this as a model for adaptation can deploy agents to do their voice bidding. Calls would get right to the point and save everyone time, and unicorns would frolic in small business lobbies whilst simultaneously spraying saved-time glitter.
Except that there is no way this is going to work well at all. Google is trying to replicate human conversation while simultaneously trying to synthesize human sociability—and the structural holes within its efforts are palpable.
First of all, it’s creepy. The examples on the Google AI blog are of calls that start with cadences that Google seems to think make the voices sound human such as, “um, hey” or “hi, uh” or other variations. The voices go in and out with instruction processing and a trained ear can hear the faux-friendly, human-like interaction interspersed with the prompt calendar and time-stamp accounting that is part of what the Google Assistant manages. It’s downright unnerving to hear humans not just engaging with the Google Assistant voice, but fulfilling its requests and instructions.
The Google voice continually saying “uh” or “um” makes it sound insecure and lacking in confidence. The Google Assistant voices are also demanding. They want. They say, “Hi, um, I would like to…” or “Hi, uh, I’m calling to ....” These are not full markers of sociability. The Google Assistant voices on each example call do not engage in any pleasantries and social markers of relationship or cooperative engagement other than to greet, stutter an “um” or “uh,” and then state their demand.
There’s no “May I please reserve a table for …” nor “May I please schedule a….” — it’s “I want” all the way down. (Note: If Google Assistant doesn’t say “Pretty Please” to a human, why should the human have to say it to the “Google Assistant”?)
Google also states that Duplex has “a self-monitoring capability, which allows it to recognize the tasks it cannot complete autonomously” and which “then signals to a human operator, who can complete the task.” Thus, the technology isn’t fully autonomous or automated. Even Google at times will need to rely upon human labor to complete certain complex requests.
This makes a strange User Experience for the person answering the phone at the business the Google Assistant is calling, for during a conversation the voice will change. They are first engaged with the “Hi, um... “ voice they heard when they picked up the phone, yet it might suddenly change into someone else entirely, with a new operator voice, personality, and their unique human personal communication style.
Unless...the only way to maintain continuity would be for Google to have human operators type the desired speech into a computer in real-time that will then be converted to the original voice that was used, which will play back to the person who answered the phone in the original “caller’s” voice. It was exhausting just typing that description, and I’m not even talking with a Google Assistant.
This has the potential to become a burdensome nightmare to small businesses that can’t afford the time or stress to step-by-step hand-hold Google Assistant towards the solution of either negotiating back and forth when an unanticipated question comes up, or needing to repeat things, or any number of things that go wrong in these types of interactions.
Google has created a mechanism that culturally behaves like a big enterprise business
As such, small business employees will now become involuntary robot coaches and nannies, helping the Google algorithms to complete their requests. Small business clerks are going to have to work through misunderstandings and clarifications and train the Google Assistant—and all of this takes time, which no business, especially a small business, has a lot of. Retail staff positions have been cut in the US and UK, and the likelihood of small businesses devoting their meager time resources training a Google Assistant, and patiently teasing out the meaning of each phone call, repeating words to clarify them, and coaxing meaning to come to fruition so that they can hang up the phone, just isn’t going to happen.
What will likely happen is that small businesses will hang up the phone, or will become successfully adept and efficient at recognizing and processing the Google Assistant calls, but will feel a bit icky in the process. Small businesses like relating to people, and no matter how much glitter you glue on an Autonomous Agent, if it isn’t properly social in the right context, it becomes creepy.
This type of artificial agent could be challenging for small businesses in other ways as well, for small businesses thrive on sociability and relationships. This is the most glaringly obvious blunder on the part of Google. Google has created a mechanism that culturally behaves like a big enterprise (impersonal, to the point, dealing with facts, process oriented), yet connects and interacts with a small business (building relationships, getting to know people, doing small favors to maintain ties). People who want a personal touch frequent small businesses.
Small businesses may have had to adjust to the world of the internet and of messaging, but in the background, they are trying to make meaning, form connections in the community, and relate socially to their customers. These touch points, in combination with other interactions, form the social relationships and bonds that enable small businesses to thrive. They will not be able to do this as fully with the Google Assistant who is now the booking agent.
With the Google Assistant, small businesses will lose a customer “touch point” at the point of appointment or reservation making, which will shift that relationship knowledge and touch point to Google. Thus, Google Assistant enables expansion of capabilities for Google and its customers, while shrinking those of the small businesses it leverages.