Human and AI Art
The unpredictable future of online shopping
Fast forward three decades, and the shopping experience has reached a new level of convenience and interactivity. Now, with conversational AI assistants like Alexa, you do not even need to lay a finger on your computer. Simply engage in a conversation with your AI assistant, and you good to go. You can instruct it to make a purchase on your behalf. Voice shopping features have truly streamlined the process, eliminating the need for extensive searching or tedious typing. As Wally Brill, a legend in conversation design, puts it: You are in the kitchen, preparing a delicious dinner with raw chicken, and at the same time, you are effortlessly adding items to your cart using voice commands. It is a whole new era of multitasking and seamless shopping experiences, as depicted in the commercials.
As fascinating as these advancements are, it is essential to maintain a critical eye on the implications of such technological progress. While voice shopping seems to offer unparalleled convenience, it also raises concerns about data privacy and security. As we rely more on AI assistants to make purchases, they gather significant amounts of personal information about our preferences and habits. This data can be valuable for targeted advertising, but it also leaves consumers vulnerable to potential data breaches or misuse. Striking a balance between innovation and safeguarding user privacy remains a crucial challenge for the future of voice shopping and conversational AI technologies.
While voice shopping falls under the umbrella of e-commerce, it significantly diverges from traditional online shopping methods. As we delve into this area, we recognize the importance of selecting appropriate research methods to understand its unique dynamics. In our recent study we wanted to get a better picture of how people voice shop. Thus, we surveyed over 300 experienced voice shoppers in the UK (in 2022) to obtain some more descriptive data firsthand.
|Which device do you use for voice shopping?
|Do you use a screen while voice shopping?
|How often do you engage in voice shopping?
|Since when you engage in voice shopping?
|Average spending per year in £ (GBP)?
|Average = 415.82 GBP
Interestingly, voice shopping seems to share conceptual similarities with decision-making in brick-and-mortar stores, where customers engage in face-to-face interactions with salespeople. We propose that customers may perceive their conversational AI as quasi-sales agents, akin to human salespeople guiding them towards making informed purchase decisions.
However, the perception of conversational AIs as shopping assistants has not received sufficient attention in research so far, despite the plethora of studies on recommendation systems. Moreover, existing findings present a paradox, showing that conversational AIs can evoke feelings of both empowerment and friendship, albeit for specific products.
In a nutshell, voice shopping’s slow adoption rate and the contrasting results in existing research prompt us to explore the possible link between purchase preferences and consumers’ perceptions of their AI assistants. By understanding these dynamics, we may be able to unlock the factors that foster or hinder voice shopping’s widespread acceptance.
Have you ever wondered why we seem to treat machines like they are almost human? It is a fascinating trend found in many studies exploring human-AI relationships. We just can not help but assign emotions and intentions to those robotic counterparts. It is like our brains are wired to make machines feel more like one of us. It is remarkable how we tend to anthropomorphize machines, attributing them with emotions and intentions as if they were one of us. This peculiar aspect has led experts to draw on psychological theories and apply them to our interactions with AI, seeking to understand our behavior and predict how we engage with these technological companions.
But, what do we exactly talk about, when talking about human-AI relationshops? Pentina, Xie, Hancock & Bailey (2023) provide an overview of consumer-machine relationships reviewing 37 peer-reviewed empirical studies. The theories used in theses studies stemed from social psychology (e.g., Bowlby’s attachment theory), communication studies (e.g., uses & gratifications paradigm), human-computer intraction (e.g., CASA paradigm) or other, like the parasocial interaction theory. These are just a few examples of the plethora of theories applied in this field. Each theory brings its own set of strengths and weaknesses to the table, enriching our understanding of human-AI relationships in distinct ways. Yet, open questions remain unsolved:
Recently we published our initial study on how humans perceive their relationship with conversational AI. Through the lens of Fiske’s relational models theory, our research revealed intriguing insights into how users establish connections with AI systems.
We identified three distinct relationship models that users adopt. First, there is the traditional master-servant relationship, where users perceive themselves as the authoritative figure in control, while the AI system serves their commands. Second, some users develop a friendship-like relationship with the AI, imbuing the interaction with a sense of camaraderie and companionship. In this scenario, the AI becomes more than just a tool; it evolves into a trusted ally, capable of offering support and understanding. Lastly, we observed a rational relationship model, wherein users treat the AI system as a somewhat equal partner. This dynamic reflects a more balanced interaction, where both parties engage in a collaborative exchange of information and decision-making.
Unraveling these diverse relationship models sheds light on the multifaceted nature of human-AI interactions, enhancing our comprehension of the evolving dynamics between users and AI systems. In a followup study we investigated the role of human-AI relationship perception in voice shopping decisions. Specifically we asked, whether the kind of relation people have with their AI, influences what kind of products they buy. In short, we found that the perception of the conversational AI as friend had the strongest predictive power for high- and low involvement products. The perception of the AI as a servant also predicted low-involvement shopping.
Indeed, to establish causal relationships and make robust claims about the nature of human-AI interaction, an experimental approach is essential. While our explorations into the various relationship models provide valuable insights, experimental studies enable us to manipulate variables and assess their impact on the interaction. By designing controlled experiments, we can systematically test different conditions and observe how they influence user behavior and their perceptions of AI systems. This allows us to identify cause-and-effect relationships and gain a deeper understanding of the underlying mechanisms driving the observed patterns.
We are currently evaluating our experiments where we tested the effect of a more emotional design on voice shopping decisions. We cannot provide results yet, however, sharing the experimental procedure where we manipulated Alexa’s output. Overall, we created four videos. Two standard shopping scenarios where a person is purchasing something over Alexa. Then we created two videos, where Alexa was more emotional from a conversational design perspective.
Example standard video: In the standard video, a person is cooking in the kitchen. The person “wakes” Alexa up telling it that the coffee machine broke down and intiates the process of buying the coffee machine via Amazon while still cooking dinner. The conversation is very close to the original; however, to be able to manipulate the output we needed clear time cuts. Thus the whole video was scripted. The human as well as the Alexa outputs were recorded. Alexa outputs were manipulated using the text to voice skill and then recorded. The videos and recordings were cut and edited using MS Clipchamp.
Example emotional Alexa video: As we wanted to explore whether a more emotional design had an effect on people, depending on how they related to the system we manipulated Alexas output. The videos were identical but we changed the wordings to make Alexa friendlier and more welcoming. In a pretest, we found that participants rated the manipulated version significantly more friendly. We relied on prior research suggestion signalling identity by using first-person pronouns (e.g., Alexa referst to itself as I or to both as we) or by signaling empathy (e.g., Oh no, I am sorry to hear that!). Furthermore, we created screenshots of the Alexa output. In the manipualted version, we furthermore inserted emojis to humanize the design.
The future of voice commerce looks promising as conversational AI users easily make purchases with simple vocal commands. Projections indicate that generative AI will further revolutionize e-commerce. However, voice shopping adoption is still limited, and the reasons for this are not fully understood. One potential factor is the perception that digital assistants lack the warmth of human sales representatives. Currently, our research is delving into the impact of emotional design on voice shopping intentions, and we will share our findings in the next 6-12 months. As we delve deeper into the realm of voice shopping and its association with emotional design, we will keep you updated with our latest findings. Stay tuned for more insights and discoveries in the coming months.
Our experts will get in contact with you!
Our experts will get in contact with you!
Further articles available here