The Intelligent Friend - The newsletter about the AI-humans relationships, based only on scientific papers.
Hello IF readers! This is the eighth issue of Nucleus, where you find insights from research papers, links to articles and useful resources of various kinds related to AI-human relationships. Enjoy this issue!
If you're a machine I don't buy
We have often talked about the prejudices we have when comparing AI or decisions made or supported by machines. But does this really impact our purchases? Spoiler: yes. This study involved 6,255 customers of a large financial service company who were randomized to receive sales calls from either chatbots or human agents under six different conditions, varying the timing of chatbot disclosure: that is, when it was actually declared that it was a chatbot.
Interestingly, when customers were not informed they were speaking to a chatbot, the AI chatbots performed as effectively as experienced human workers and far better than inexperienced workers in securing purchases. However, revealing the chatbot's identity before the conversation led to a dramatic 79.7% reduction in purchase rates and significantly shorter call lengths, suggesting that early disclosure negatively impacts customer perceptions of the chatbot's knowledge.
However, there is an insight that in my opinion is very useful for different types of business: delaying the disclosure until after the conversation or after the purchase decision helped mitigate the negative effects, with purchase rates similar to the no-disclosure condition. Additionally, customers with prior experience using AI were less affected by early disclosure.
Title: Frontiers: Machines vs. Humans: The Impact of Artificial Intelligence Chatbot Disclosure on Customer Purchases. Author(s): Luo et al. Year: 2019. Journal: Marketing Science. Link.
A concrete impact on self-depression
AI is growing a lot from the point of view of relational dynamics in adoption by consumers, especially because there are problems whereby these technologies are seen as potential 'actors' of negative effects by consumers. An impact case is brought by this paper by Martinengo, Lum and Car (2022): the investigated how chatbots can support individuals managing depression. In particular, The research analyzed nine different mental health chatbots by engaging them with standardized personas representing users with depression.
I summarize the most relevant insights in my opinion from this study which I really liked and which focused attention on an important and delicate topic:
Chatbots were able to respond suitably and remember previous interactions in many cases. Empathy was conveyed effectively through supportive statements, understanding user feelings, congratulating on task completion, and offering support during distressing times;
The chatbots provided guidance for various mood-boosting activities based on several psychotherapeutic approaches, helping users engage in structured exercises to improve their mood.
Mood monitoring was typically conducted using predefined options, such as words or emojis, with some chatbots administering standardized questionnaires for depression screening.
As specified in the 'Discussion' section of the paper, this study was particularly innovative because, also methodologically, it was the first to implement an analysis of dialogues between users and chatbots.
Title: Evaluation of chatbot-delivered interventions for self-management of depression: Content analysis. Author(s): Martinengo, Lum and Car. Year: 2022. Journal: Journal of Affective Disorders. Link.
The issue(s) of the week
Among the many analyses regarding Apple Intelligence, I particularly liked this one, by
of .There is an increasingly need to be 'Polimatic', in some ways, in my opinion, yes. And to show it is this wonderful piece from last year that I read recently by
who answers a rather burning question.I also take this opportunity to remind you that this week my first personal 'essay' was released here on Substack, in which I try to answer a question that I have often been asked: why do I like reading papers so much? With references to Feynman and machachi groups, I have tried to give my perspective which I hope you will find interesting.
Not a machine doctor, please!
Today's first paper tried to highlight the reactions in purchasing situations to the disclosure of an interaction with a human vs. a chatbot. This study has instead shifted the research focus to another theme: healthcare, where, if on the one hand AI can be of great support, on the other it is fundamental to understand what the factors may be that affect acceptance or opposition of these tools by patients.
In this study, researchers explored exactly the reluctance of consumers to engage with medical AI compared to human providers. During the experiments, participants were asked to choose between human and AI providers for various medical procedures, including skin cancer screening, stress diagnosis, and coronary bypass surgery recommendations.
Keep reading with a 7-day free trial
Subscribe to The Intelligent Friend to keep reading this post and get 7 days of free access to the full post archives.