“Voice has the power to transform computing, by providing a natural means of interaction” – The Economist (2017).
People think that artificial intelligence or algorithms are a new technology that has only just been invented. In reality, algorithms have existed for many, many years already. The only difference is, there has been huge improvements in the field of artificial intelligence in recent years and large companies began to incorporate that technology in their products. Due to the improvements and the rising popularity of those products, the use of artificial intelligence, digital voice assistants, smart interfaces, smart home devices, intelligent agents (all names that are used to describe the same type of technology) has increased. Therefore, understanding how people interact and communicate with digital voice assistants can provide useful insights and information for marketers as well as graduate marketing students.
The aim of this paper is to collect data on how different people interact with a digital voice assistant and how the results can help marketers to understand and identify opportunities to use that knowledge in their marketing strategy.
With Amazon, Google and Apple each launching a voice-controlled speaker within the last three years, the market for smart and voice-controlled devices and the use of artificial intelligence (AI) continues to grow (Mintel, 2017). This is a huge shift in the computing industry, since now the consumers are able to use a more convenient and natural way (The Economist, 2017) (speech) to do their day-to-day tasks, such as setting alarms, creating shopping lists or ordering food (Stirrat, 2016).
Statistics show that the percentages of use of voice searches and commands within different age groups is relatively high (see Figure 1). However, according to a survey by YouGov (Feldmann, 2018) on how smart speakers are used, only 9% said that they use their devices to order products online. The reason for such a low percentage is that people have trust issues ordering products online by just using their voice. They are used to seeing the things they order and purchase on screens or live in-stores (Armstrong, 2018a & 2018b). However, a study by OC&C Strategy Consultants (2018) shows that purchases through voice activated smart speakers in the UK are currently worth £0.2bn and are predicted to increase to £3.5bn by 2022. Through the huge increase of in-home AIs, businesses are aiming to keep up with this new technological trend and to reach that predicted number of sales (Gingiss, 2018). In order to achieve that, marketers need to understand how people interact with artificial intelligence and smart speakers so that they can adapt their marketing strategies in order to influence their consumer behaviour and to drive sales.
There is a plethora of research and studies about human-computer interactions and even newer academic articles about artificial intelligence (AI) and how humans interact with it. Additionally, a lot of secondary data can be found regarding the public's online consumer behaviour. The identified concepts and theories can all be linked to both the human-computer interactions and the online consumer behaviour and will therefore be used as a theoretical framework for the dissertation. However, there is no further research on how humans interact with artificial intelligence, specifically voice activated AI devices, and how those interactions can influence the online consumer behaviour.
Related work on human-computer interaction
With the continuous increase of new technologies, more research and investigative studies on how people use and interact with those technologies, have been conducted in recent years.
According to the theory of social response (Moon, 2000) people treat computers and artificial intelligence as social actors and interact with them as they would with another human being. Although research shows and proves that computers do not have any human characteristics and that people are fully aware of that, they still apply social cues such as reciprocity and self-disclosure to the computers mindlessly (Nass & Moon, 2000). Chattaraman et.al. (2018) state that the psychological phenomenon of anthropomorphism can also be found in human-computer interactions since people tend to give computers as well as AIs personalities and human qualities (Edwards et.al., 2018). Computers with speech capabilities play a major role in those interactions. Voice is an important factor in human interactions since it is the most natural way of communication and therefore a key in social responses between computers and humans (Huang & Lin, 2011). An experiment conducted by Lee (2010) shows that computers with synthesised speech come across as more positive, believable and socially attracting. According to Edwards et.al. (2018) synthesised voices are more influential especially when they are similar to or match the user's personality. This could be led back to the theory of the uncanny valley which states that if objects are too human-like, people tend to have high expectations on the capabilities of those objects. Since they are not human, they cannot fulfil the expected tasks and that leads to disappointment of the user. If something is less human-like people have lower expectations and can therefore not be disappointed and feel more positive about it (Gray & Wegner, 2012).
Related work on online consumer behaviour
A lot of secondary research can be found for traditional offline consumer behaviour but with the popularity of online shopping, more and more studies have now also focused on the online consumer behaviour and how it differentiates from the traditional ones.
‘Every user is now also a computer user' (Koufaris, 2002, p. 205). This implicates that users accepted computer technology and know how to use it. The Technology Acceptance Model focuses on two dimensions: usefulness and ease of use. Usefulness is considered by people when they believe that the technology will actually help them fulfilling tasks (Davis, 1992). But before they try the technology they consider the ease of use and their own personal capabilities of performing with the particular technology (Dennis et.al., 2009). This ease of use is also closely linked with someone‘s self-efficacy and personal attitudes towards a certain behaviour which are parts of the theory of planned behaviour (Akroush & Al-Debei, 2015). For instance, attitudes towards online shopping influence the behaviour: the more positive the attitude towards online shopping is, the more likely the individual will actually purchase something online (Hsu, Chuang & Hsu, 2014). But even if the technology has been accepted, there are still several uncertainties concerning internet and technology (especially in an online shopping environment) such as data protection, anonymity, unacceptable quality and the risk of not receiving the ordered product (Salo & Karjaluoto, 2007). Those uncertainties can only be eliminated by consumer trust which can be achieved through previous experiences and following satisfaction (Kenning, 2008). Regarding human-human interactions, studies show that trust is an important part of social relationships as it deals with the risks of unachieved expectations in social actions (Farini, 2012). In order to build online consumer trust, big online firms for example included social embedding in their websites and capitalise from the positive recommendations and positive word-of-mouth of others (Bourlakis, Papagiannidis & Fox, 2008).
Although there are concerns about risk and security on the internet, there are still various motivators for people to shop online. The two most common ones are saving time and effort through convenience and fastness (Akroush & Al-Debei, 2015).
Those concepts and theories build the conceptual toolkit that will be used and expanded upon for the dissertation.
Alongside the secondary research which includes journal articles on human-computer interactions and online consumer behaviour, and textbooks, the primary research will be conducted through an observation. The purpose of the observation is the collection of qualitative data to develop an understanding as to how people interact with artificial intelligence in order to answer the research question.
The research will use an interpretive methodology as the data that has been collected through observations needs to be interpreted since it is ‘based on human situations and interactions, which are always a bit different, non-replicable, and non-identical' (Wisker, 2009).
It will be an overt observation since the participants will be informed about the whole process (Malhorta, 2017). With observation research methods, the observer can have a certain degree of control by imitating real life in order to see how the participants behave in such. Regarding the proposed idea, it is the ideal way to collect data as observational data will be credible and faithful (McNeill & Chapman, 2005).
The observation will take place in the consumer behaviour laboratory at the School for Business and Technology. 20 people will be chosen using various demographics and sociographics such as age, gender, nationality, culture, occupation, experience with AI, etc. Artificial intelligence can be used by everyone regardless their market segmentation, hence, there will not be any limitations in that. Since the sampling will only consist of 20 people, this research is exploratory (Guthrie, 2010) to give a first approach. Each of them will be given a piece of paper with 10-20 different tasks, each paper will have the exact same tasks. They will be sent to the consumer behaviour lab on their own and ask to fulfil those tasks by using the provided Amazon Echo product. The tasks will include questions like ‘Order the product xx from Amazon online with Alexa'. The questions will be clearly formulated so that there will not be any confusion. Regarding the time that will be given to each participant, there will not be a time limit. In order to collect as much data as possible, the participants will have as much time as they need to fulfil all the tasks. The consumer behaviour lab includes screens, microphones and cameras in order to watch what happens in there from the outside. For observational purposes the participants inside the laboratory will be filmed and recorded while interacting with the digital assistant. After the observation, there will be a short, semi structured interview with already prepared but mainly open questions. The participants will answer questions like how they felt during the interaction with “Alexa”, if it was easy for them to interact, did they trust “Alexa” especially with the online order, did they enjoy it or did they find it uncomfortable, and their overall opinion on digital home assistants. These interviews will provide a deeper insight in each individuals‘ behaviour and emotions towards AI (Rugg & Petre, 2007).
The recorded interviews will be transcribed and analysed for statements and similar results and the observation data will be coded in order to classify behaviours (McNeill & Chapman, 2005).
- Amazon Echo only works when it is connected to a working WiFi connection. Since the laboratory is on university ground, the device cannot easily be connected with the internet connection. But the IT team of the university will be included in finding a solution for that.
- Since conducting an observation which will take a lot of time and is exploratory, the number of participants cannot be very high (Silver & Wrenn, 2013). Therefore, 20 people would be considered as appropriate.
- Regarding the tasks Alexa will have to fulfil during the observation (especially the online order task) it is necessary that the participants really buy something which needs to be paid for. But research on Amazon's cheapest products could be the solution for that problem.
- With observational research methods there is always the problem that if the participant knows he/she are being observed they might not act naturally (Malhorta, 2017).
In order to get the best data, the participants will be recorded and filmed. Of course, this is intruding their privacy and therefore they will be informed about that before they agree to participate in the research process. The participation will be entirely voluntarily (Denscombe, 2010). If they agree to the participation, they will get a written form with a brief summary of the process, the right to withdraw, the confidentiality and security of the individual‘s data and the signatures of both, the researcher and the participant. The form will also state that the filmed and recorded material and the answers to the interview questions will be kept anonymous and treated as confidential. They will not be used for any other purpose than that for the dissertation research.
Also, there are existing doubts regarding digital personal assistants and their security and privacy safety. All information like credit card details etc. that need to be set up with the device, will be the researchers' own details, never the ones from the research subjects.
Additionally, the tasks that will be given to the participants will not include any inappropriate or offensive topics, but in order to avoid any claims, the form will also include a statement about if anything comes across as offensive to any individual that that was not intended.
...(download the rest of the essay above)