Author: Daniel Hyunwoo Lee, Graphics: Nina Tagliabue

THE BRB BOTTOMLINE:

Virtual assistant software is now considered one of the most sophisticated and fastest-growing technologies of our time. Becoming more prevalent than ever before, the tech giants have jumped on the bandwagon to take the lead in this promising market. Whether the existence of digital assistants helps the economy, however, is a different question.

The Rise of Digital Assistants

Also known as digital personal assistants (DPAs), virtual assistant software is an intelligent automated system that communicates with the user through interactive dialogues while applying third-party services to obtain information with the goal of making users’ lives more convenient.

As with DPAs such as Google Home and Amazon Echo, products in the voice technology market seek to make the lives of users simpler by means of speech search. According to a study published by Google on voice technology, speech search is used most frequently to find travel routes and target destinations, to dictate text and notes, and to dial phone contacts. This shows that it is particularly used in moments in which the users cannot type or doing so would be dangerous, for instance, while driving a car. Due to the convenience they provide, more than 1 billion regularly used a digital assistant with speech control or speech search already in 2019 according to statista.com, and this number is expected to grow up to an astounding 1.8 billion in 2021.

Figure 1: Number of Users Per Year Regularly Utilizing Digital Assistants

Along with direct benefits to the consumers such as a more convenient lifestyle, on an economic scale, DPAs have been on the rise as they reduce transaction costs and boost the e-commerce industry. Besides its benefits, however, the adoption of DPAs means the economy has to adopt its downsides as well.

The Economic Risks of Adopting Digital Assistants

According to “The Brave New World of Digital Personal Assistants” by German Economist Oliver Budzinski, the potential economic challenges resulting from the adoption of digital assistants can be categorized into five directions—“biased services, market power on the DPA market and economic dependence on a dominant DPA, potential leveraging of DPA suppliers’ market power into neighboring markets, personalized data (ab)use and privacy, and media bias and manipulation of public opinion” (Budzinski 1). This article mainly explores the first two directions.

Biased Services, Loss of Purchasing Behavior

First, the biased services provided by DPAs alter consumers’ purchasing behaviors entirely. This is because, while traditional e-commerce still leaves it to the consumers to search and read information on their own, voice-controlled DPAs can easily become an information gatekeeper for inattentive users. Under this situation, DPAs are “likely to be employed to induce their customers’ purchasing behavior according to the interests of their suppliers” (13). In the case that DPAs recommend products of their business partners or subsidiaries, naive consumers may simply rely on their virtual butlers’ suggestions or even relinquish their decision rights to the DPAs without further concern. Moreover, imperfect information may prevent even more rational consumers from exercising sufficient control over their digital agents, which also enables DPA suppliers to make DPAs suit their business interests at the expense of the users’ welfare.

Since DPAs function based upon individualized data and algorithms, the DPA’s hidden agenda may be difficult to detect—this may be exacerbated if users get locked in to the services of a certain DPA. For instance, DPA suppliers may block the transferability of data among different assistants for strategic reasons so that users must spend significant additional time and effort to train a new assistant. Such incompatibilities among different DPAs increase switching costs and, thus, cause lock-ins and overdependence on one supplier due to the low interoperability of the product. Thus, switching to another DPA becomes disincentivized.

Dominant Suppliers, Lack of Choice

This leads to the following argument that the increasing use of DPAs also involves a high chance of creating an oligopolistic, or even monopolistic, structure in the market highly unfavorable to the consumers in the long run. The DPA market has the same features like many other digital platforms—in particular, the main competitors in the current DPA market is a duopoly market (Amazon and Google being the two) with various fringe competitors. The incumbent players build strategies to integrate isolated markets, such as the development of smart speakers or the integration of DPAs in cars.

The contestability of the market is hampered by the presence of barriers to entry and is supported by network effects, supplier-side economies of size, incompatibility among platforms, switching costs, and intellectual property law, which is associated with the scale of research and development costs. The DPA market displaying a rather narrow oligopoly structure in the coming future is not so problematic as long as the competition among the oligopolists remains intense. Notwithstanding, if the number of relevant competitors becomes narrower, “incentives to enter into a collusive equilibrium increase and DPA suppliers may coordinate their behavior by means of tacit collusion or cartel-like structures instead of competing with each other at the cost of consumer welfare” (15).

Furthermore, if the market tips into a monopoly-like structure with one dominant supplier, then the dominant DPA can exploit 1) the dependency of consumers by increasing prices and/or reducing service quality, as well as 2) the economic dependency of third-party suppliers, who require access to the DPA (and its search, recommendation, and decision-making services) in order to have access to a relevant number/share of consumers. If one DPA becomes an essential gatekeeper—Amazon and its DPA Alexa, for example—it will force unfair conditions upon them.

Other Known Dangers

In 2017, channel CW6 in San Diego, CA, aired a news segment about the vulnerabilities of Amazon Echo speakers (equipped with the Alexa virtual assistant).The system is unable to differentiate people by voice, the show hosts explained, which means Alexa will follow the orders of anybody who is around. As a result, little kids started making unplanned online purchases, not knowing the difference between asking their parents to give them a snack and asking Alexa to give them a toy. Then one of the hosts said on air: “Alexa, order me a dollhouse.” The complaints began rolling in. People all over San Diego reported spontaneous purchases of dollhouses made by their voice assistants—Alexa heard a line uttered on television as a command and swiftly fulfilled it.

Furthermore, the voices of Siri, Cortana, Alexa, and Google Assistant are all female by default (already reinforcing gender roles), with an accent close to the standard Midland American English used by newscasters in the U.S. Nobody can fault the clarity of speech of these DPAs, but it seems that some might be best served by them if they sound similar. If someone has a strong accent (for instance, Chinese, Irish, and Scottish), they may have to pick up the phone and order that pizza themselves as their DPA might not understand them. Certainly, it is difficult to create robust speech recognition, especially if there is little data to train from—as is the case for minorities. However, as speech recognition is incorporated into more systems that we encounter daily, imagine the delays and frustrations one might go through if their accent is not understood.

In addition to trying to include understanding of diverse accents, engineers need to be aware of other biases in the data that chatbots may learn. For example, Tay, a Twitter chatbot from Microsoft, was shut down after it was used by the so-called ‘trolls’ to spread racist tweets. More recently, two chatbots in China were also shut down after they expressed disdain for communism. Speech recognition and language generation without restrictions seem to always end in misbehavior.

These concerns point to a need for engineers to be aware of the social context in which they create and release language-enabled devices. They must consider not only the intended uses but how they can be misused—by the owners, oblivious or malicious outsiders, and by law enforcement.

The Future of Digital Assistants

In the near future, there will be a development of an agent that will get rid of the use of mobile applications. The agent will be an interface for the individual consumer’s personality, values, and budget that automatically interacts with any other piece of software or human in the world. The agent will take time to gather information and understand its owner and their values, then will be able to execute the user’s will effortlessly and anonymously on the behalf of the user.

Such a product will certainly make our lives convenient, but considering the devastating economic impacts it can potentially have through taking away customer independence and the free market structure, it should be created not for the purpose of monopolizing the market, but rather to give power back to the consumers.

Take-Home Points

DPAs not only have the potential to provide biased services and lead to a loss of purchasing power but also to decrease freedom in the market through fostering an unfavorable market structure for consumers.
Use Alexa at the sake of economic disruption—do not let your assistant overtake you.

Your Alexa Could Be the Cause of Economic Disruption