Amazon Go and the Erosion of Supermarket Sociability

Invited post by: Sally A. Applin (@AnthroPunk on Twitter)[1]

I recently finished my Ph.D. As a present, a friend of mine gave me a hand. Not help, which he had done during the process, but rather a battery-powered automated hand, cut off at the wrist, similar to that of Thing, the Addams Family’s servant from TV and film. In part of my thesis, and my research on automation, I’ve looked to Thing as a metaphor for IoT software automation. Thing, on TV, is a trusted friend who builds relationships with family members and can negotiate with others on their behalf. In fiction, and the representation of fiction, Thing works beautifully and embodies what a smart agent could be. It is aware of its surroundings, it builds trust. It connects people. Thing is a keeper of local knowledge. The Applin and Fischer (2013) Thing agent, is a software construct using deontic logic to encourage and support human agency, building trust in a relationship based context.  The hand my friend gave me moved on a fixed path for several seconds, and then stopped until its button was pushed again. It looked like Thing, but it was only a physical representation, a simulation of physical form. In automation, data collection is not the same as building relationships, and community knowledge cannot easily be derived from quantitative Big Data. This is one of the more serious problems with Amazon Go.

Amazon Go is a grocery store concept that allows people who have activated the Amazon Go app on their mobile phone, to walk through an “authentication” turnstile into an Amazon Go supermarket. Once inside, people can “grab” what groceries they want or need, and walk out the door, without needing to check out, because Amazon’s “computer vision, sensor fusion, and deep learning” will calculate what people take, and charge them accordingly via the app. Amazon Go has a video on their website that explains all of this, and shows people “grabbing and going” with their groceries, stuffing them into bags or just holding onto them, and walking out. In the Amazon Go video, no one is shown talking to each other.

On the surface, Amazon Go may seem ideal to people who need food and lack time. Supermarkets are a mess. The current market I go to has long checkout lines as well as a huge navigation problem. Mostly the navigation problems are due to online shopping service workers, who are tracked and clocked and thus rush carts around the store to fulfill orders amidst others operating at a different pace (similar to what we will start to see as more autonomous vehicles join us in more numbers on the roads). This may not be any different for Amazon Go. People might prefer a store with no check out, but might take their time shopping and browsing and block progress for those harder core grabbers and goers. Amazon Go seems to be turning shoppers into mini versions of the Amazon order picking robots, clearly trying to leverage their warehouse technology to apply to grocery shopping. However, Amazon doesn’t seem interested in building trusted relationships with shoppers as Thing—if Thing were real. In Amazon’s world, shoppers are tracked, and measured and their data is reported back to Amazon, who isn’t sharing it for anyone’s benefit but Amazon’s. People’s bodily motion, response rate, etc. will be tracked as well, which is a huge forfeiture of privacy.

Amazon Go’s focus on “grab and go” seems to be about time savings. This is not unexpected, as Amazon has demonstrated more than a little Quantitative focus (bias). Amazon’s rush to quantify and automate has seemingly depleted its ability to examine both its impact on society, and its own limitations. Amazon does not demonstrate prioritizing people, or the personal side of automation and agency. The main metric that Amazon has focused on is saving people time, which has somehow evolved to be the thing to save, based on early manufacturing models of assembly line efficiency. It isn’t clear if Amazon Go is considering what the “grab and go” action does—does it train behavior on an aggregate scale? Does it have the potential to create long term damage to communities, as sociability is removed from the shopping experience?

Grabbing and going seems to be a revisited human theme in current society. It’s popular with toddlers, and has a history for adults as well. Through the ages, humans have plundered and looted as a “reward” of war; to make ends meet; and/or to enrich their own situation at the expense of those who are deemed to have more than enough. What is new is how it has resurged. A current trend in burglary is “grab and go.” Certain stores are targeted over others, and Apple stores are particularly vulnerable. The Burlingame, CA Apple store was recently robbed three times in this way, twice within four days. In about 10 seconds, 3 to 5 robbers entered an Apple store, untethered and grabbed devices on the tables closest to the front door, and ran out. In New York, a man “grabbed” an 86 pound bucket of gold flakes worth $1.6 million dollars off the back of a truck, and walked down 48th Street with it. “Grab and Go” is clearly a time saving strategy.

What robbers are doing with “Grab and Go” in each case is exploiting a vulnerability in a system—preying on predictable social interaction, or finding a loophole where there is no local surveillance. (We’ve seen other exploits of the same type , such as what happened on 9/11.) Exploiting vulnerabilities in a system allows for certain types of innovation, as well as exploitation. In innovation, Amazon Go is enabling human agency for ease of product procurement, while saving time for individuals, however this may be coming at the expense of the community group in a broader way, as it simultaneously exploits individuals for their behavior patterns.

Marketplaces throughout human history have been a vehicle for trading commodities, and one of their main functions has been to create a venue for people to exchange information and ideas, and to socialize. This social exchange is what enables us to cooperate with each other in broader ways outside of the marketplace. Additionally, because vendors interact with so many people in the community, those working in the marketplace are the keepers of community knowledge—the understanding in a holistic (or close to) way of the health of the community writ large.

In moving the marketplace to a commodified and quantified framework, Amazon has removed the sociability of the marketplace, which is a vehicle for both community cooperation, and the qualitative data of a community’s well-being, replacing it with its sole ownership of the aggregate of individual quantitative data. Amazon Go will be activated without any consideration of its impact to the community it will be deployed upon. Furthermore, the community knowledge that comes from trusted human relationships will now be seemingly “created” by analyzing Big Data algorithms, which will not produce anywhere near the same result—without broader contextual knowledge. Community knowledge is local and specific. It is also critical to the way that people cooperate and survive in their communities.

The power of Amazon Go is that it is creating a marketplace for people with time constraints and mobile phones, which is just about everyone. The problem is that in doing so, Amazon Go will be contributing to changing community structure even further than the advent and broad usage of mobile phones, and soon to be other autonomous based services. As we’ve migrated away from our local locales, relying more and more on our networks for community knowledge, we too, “grab and go” information and data, applying it only to ourselves and our lives. The problems arise when we need to communicate and cooperate in our local locale, but have a network that is distant and distributed, or based solely on quantitative data.

[1] Sally A. Applin earned her Ph.D. in Anthropology at the University of Kent at Canterbury, UK, working with the Centre for Social Anthropology and Computing (CSAC) where she researches the changing relationship between humans and algorithms, the impact of technology on culture, Maker culture, leading technologies, and the outcomes of network complexities as modeled by PolySocial Reality (PoSR). Sally holds a Masters degree from the graduate Interactive Telecommunications Program at NYU (ITP), and a BA in Conceptual Design from SFSU. Sally has had a career in the science museum design, computer software, telecommunications, innovation, insight, and product design/definition industries working as a Senior Researcher, and Senior Consultant. Sally is an Associate Editor of the IEEE Technology and Society Magazine, and Associate Editor of the IEEE Consumer Electronics Magazine (Societal Impacts Section), a member of IoT Council (a think tank for the Internet of Things (IoT)), and a board member of the Edward H. and Rosamond B. Spicer Foundation.



Ryan Anderson is a cultural and environmental anthropologist. His current research focuses on coastal conservation, sustainability, and development in the Californias. He also writes about politics, economics, and media. You can reach him at ryan AT savageminds dot org or @anthropologia on twitter.

7 thoughts on “Amazon Go and the Erosion of Supermarket Sociability

  1. Hi Sally, I really enjoyed your post, particularly the idea of grab and go theft. I’m curious, do you have any sense or prediction of what sort, if any, local sociability will pop with grab and go in stores, or is it alienation and dis-embedding all the way down? Maybe a related question–if grab and go’s idea is to save people time, where does that time go?

  2. Hi Daniel, good questions. Allow me to contribute some observations concerning convenience stores in Japan. First, convenience stores are, by design, a grab and go business model that makes it simple for strangers to pop in, find what they want and check out quickly with a minimum of social interaction. Only vending machines provide less. In one of the studies from which I took the data for my book on Japanese consumers, the Japanese researchers contrast convenience stories with neighborhood “snacks,” small local pubs with an established clientele. The convenience store offers variety, limited by floor and shelf space. The “snack” offers a severely limited menu but the offerings are tailored to its clientele. Thus, for example, a regular need only sit down at the bar to have his “usual” set before him. Amazon’s online sales model is a combination of these approaches, combining a vastly extended variety of product with use of big data to more and more closely anticipate each regular shopper’s needs.

    So far, however, I have only been describing concepts. What of on the ground realities? There is likely to be, I suggest, a strong difference between a neighborhood convenience store and a convenience store located in a major transport hub or shopping complex. I think, on the one hand, of the7-11 at the foot of the hill on which our apartment is located. It was once a local liquor shop. The owners are long-time residents as are many, if not most, of their customers. Local teenagers hang out in front of the store. I think, on the other hand, of the 7-11 at the top of the escalator where I sometimes exit Yokohama Station. This store caters to an ongoing stream of travelers that becomes a crush at peak commuting hours. Personal interaction is radically minimal.

    Convenience stores are subject to grab and go theft. The Amazon grab and go brick and mortar shopping model will, I suspect, be much less so. Anyone attempting to leave a store without having paid for whatever products they are holding will be instantly detected by the sensors they must pass before the door will open for them. What further security measures will be taken, I do not know; but there will be no slipping out while the clerks are distracted by other customers.

    Hope this is helpful.

  3. Hi, Sally. I also enjoyed the post. The topic of how different retail scenarios affect communities and other social relationships is a fascinating one. That there is a gap between a small, local market where merchants and customers know each other and a grab and go world in which consumers are conceived as anonymous monads and the merchant’s primary goal is to increase profit through greater efficiency is undeniable. When, however, I ask myself what fills that gap, I see all sorts of possibilities. In Japan, where I live and work, retail has undergone a series of transformations since the end of WWII. First, a shopper’s world divided between high-end department stores and local shopping streets was disrupted by the introduction of supermarkets. Then came fast food and convenience stores, the latter made possible by the introduction of point of sale (POS) systems that made it possible to track what was and wasn’t selling on a daily basis. Now on-line shopping has been added to the mix. Recent years have seen a resurgence of co-ops and restaurants that deliver to the customer’s door. Each of these innovations has taken place in a particular context: supermarkets rising with Japan’s “New Middle Class,” nuclear families living in suburbs or massive condominium apartment blocks; fast food and convenience stores catered to a growing number of young singles, though in neighborhoods like mine the number of again customers is growing; on-line shopping appeals not only to young families with working mothers but also to older people (in Japan 25% of the population is now 65 or older); and restaurants that deliver are convenient for all sorts of people, elderly folks for whom going out can be literally painful, as well as office workers putting in overtime. How operations like Amazon Go will fit into this mix is a a timely and important topic.

  4. Hi, thank you both for your thoughtful comments. I will address Daniel’s first: You asked “…do you have any sense or prediction of what sort, if any, local sociability will pop with grab and go in stores, or is it alienation and dis-embedding all the way down? Maybe a related question–if grab and go’s idea is to save people time, where does that time go?” – My Reply: I think that some initial sociability will happen as people are “amazed” or “wowed” by not having to check out in the usual way. There could also be sociability around problems or issues as the technology is flawed. Amazon has recently announced that there will be humans staff initially to address issues (e.g. things that go wrong when the algorithms break, which they will). That said, once these things are resolved, it may be “alienation an dis-embedding (OF THE LOCAL LOCALE) ” all the way down as people are very well social and connected, just not to the people next to them in their communities. They have strong and rich environments, but these are in the form of distributed personal networks online. If “grab and go’s” idea is to save people time, the time is going towards other time deficits we have now as a result of having to be our own admins for services that were previously done by humans in the community, and which have now gone online. See also Applin and Fischer 2013: “Asynchronous Adaptations to Complex Social Interactions”

  5. John, thank you for your interesting comments and for sharing your research on Japan’s Combini and retail landscape. The main difference I see between Japan’s “close to Grab and Go” and Amazon’s Grab and Go is that Japan is a mostly homogeneous culture, and that the US, where I am writing about (but other cities can and do fall into this model), is an extremely heterogeneous society. As such, there is less shared cultural and community knowledge to begin with, which makes the US model more fragmented in local locales. In Japan, because there is higher homogeneity, I would argue that shifting to more automation may be less of an issue as long as the culture remains homogeneous and other conventions in Japanese communities can sustain cultural and local community knowledge.

  6. Anthropunk, I hear where you are coming from, classic stereotypes about more and less homogeneous cultures. But given the difficulty of making sweeping generalizations about a nation-state geographically the size of California with a population of 127 million people who live in places that range from depopulated rural villages to the Kanto Plain, one of the largest and most sophisticated metropolitan areas in the world. I have mentioned the difference between my neighborhood 7-11, the reincarnation of what was once the neighborhood liquor store on a street lined with local businesses, and the 7-11 at Yokohama station whose customers are commuters who only rarely encounter someone they know. Both are in Japan. Here is precisely where ethnography, local knowledge and thick description become essential.

  7. I have traveled extensively in Japan and am aware of the usual stereotypes, but am also aware that there are strict immigration rules that keep for the most part, a similar (well more than there are here in Silicon Valley) culture. That matters! This piece laments that ethnography, local knowledge, and “thick” description may be a casualty of automation, by merely stating, that in a store with no human anchors on a daily basis (although Amazon has said they will have them for a bit), there is a community knowledge that goes missing, since part of transactions involve some form of social exchange. When that is removed (no matter how inconsequential they do add up), it impacts knowledge of a locale, and by extension I argue that this is amplified in heterogeneous cultures because there is more fragmentation of cultural “glue.” Your locale may be different and indeed, you also have not yet studied these “grab and go” environments because they do not exist yet without human staffing, except for this new Amazon model. However, since this piece was published, Amazon has already found out they need to add human staff, thus temporarily preserving cultural community knowledge. Huzzah!

Comments are closed.