The Core of AI-Bility
Let’s have some information about our project in a question-like style ?
Background: Why did you apply for this project? What are the needs you plan to address?
We observe the abundance of smart toys, adaptive learning applications, and digital assistants for schoolchildren on the market. These products are artificial intelligence (AI) based conversational agents that can communicate using natural language. The current pandemic challenge probably plays an important role in promoting their adoption. The truth is, however inconvenient, that we still do not know much about how schoolchildren harness these AI-based conversational agents for their benefit. Because of the way they are designed (i.e., real touchable physique versus digital character), schoolchildren may have different kinds of interaction and experiences with them. Moreover, they may be perceived differently because of their appearances (i.e., pet-like and human-like characteristics).
This project addresses a discussion that is likely to become more glaring in the next years, due to the increasing adoption of AI-based conversational agents for learning and leisure activities. We focus on 11 to 13 years old schoolchildren in Liechtenstein, Germany, and France. In this age group, they begin to learn abstract reasoning (Jean Piaget’s concreate and formal operational stage) and develop belief in their own ability to solve tasks together with a sense of identity in relation to their social others (Erik Erikson’s stages 4 and 5 of psychosocial development). We also take into account the concerns about the digital divide between schoolchildren and their caregivers, and a cultural gap between digital natives and digital immigrants. By understanding how schoolchildren perceive and interact with conversational agents, we can equip parents and teachers with hands-on know-how in guiding schoolchildren to develop a mindful and healthy interaction with AI-based conversational agents.
Objectives: What do you want to achieve by implementing the project?
Responding to the problematisation, we pursue two main objectives in this project:
- To explore and understand how schoolchildren interact with different types of AI-based conversational agents and how they perceive these conversational agents compared to their existing social others (such as family, friends, and teachers). This objective is addressed in work packages 1 (conceptual development) as well as 2 and 3 (field study).
- To equip schoolchildren as digital natives and their caregivers with hands-on knowledge in dealing with the rapid advancement of smart technologies, especially those that appear to be social and human-like in either appearance or intelligence. This objective is addressed in work package 4 (storytelling blog).
Our first objective is about understanding schoolchildren’s digital readiness and capacity in harnessing new digital technologies. Based on this understanding, we aim to equip schoolchildren, their parents, and their teachers with hands-on knowledge that, again,
promotes digital readiness and capacity in interacting with new digital technologies. Ultimately, we wish to join ongoing
European-wide initiatives that aim to provide a better and safer digital space for children. Finally, we also learn
from the European Union Agency for Fundamental Rights (FRA)’s (https://fra.europa.eu/en/project/2018/artificialintelligence- big-data-and-fundamental-rights) discussion about artificial intelligence and the fundamental rights. In
comparison to other ongoing Erasmus+ projects on the subject matter, our focus is on the social and cognitive
consequences of AI-enabled CA use for schoolchildren and not on how to teach AI to children.
Implementation: What activities are you going to implement?
We divide our main project activities into four work packages (WPs):
- WP 1 – Conceptual development. We will analyze available AI-based conversational agents on the market that specifically target schoolchildren and dig deeper into their design purpose, working mechanisms, and potential pros and contras. We will develop a typology of these conversational agents.
- WP 2 – First field study. We will interview 11-13 years old schoolchildren to explore how they perceive and interact with AI-based conversational agents.
- WP 3 – Second field study. We will investigate the implications of the new learning strategies that schoolchildren develop when using AI-based conversational agents, compared to the more traditional learning strategies.
- WP 4 – Dissemination to a wide audience. We will share the insights directly with participating parents and teachers. We will also use the storytelling approach in disseminating them on a dedicated blog.
We will also use the storytelling approach in disseminating them on a dedicated blog. Each work
package contributes to the achievement of our project objectives.
Results: What project results and other outcomes do you expect your project to have?
By understanding how schoolchildren perceive and interact with conversational agents, parents and teachers will know how to guide them in developing a more mindful and healthy interaction. Moreover, this project addresses a discussion that is likely to become more glaring in the next years, due to the increasing adoption of AI-based conversational agents for learning and leisure activities. Finally, this project embraces both the bright side and the dark side of technology. We foresee four main project results:
- Result – A typology of conversational agents for schoolchildren. We will analyse available AI-based conversational agents on the market that specifically target schoolchildren and dig deeper into their design purpose, working mechanisms, and potential pros and contras. We will develop a typology of these conversational agents.
- Result – Insights on how schoolchildren perceive and interact with AI-based conversational agents. The setting of the exploratory field study We will conduct focus groups in schools (highest standard of education) in Liechtenstein, Germany, and France with schoolchildren aged 11-13. With the help of semi-structured interviews, we will explore how school kids interact with AI-based conversational agents. We will focus on concepts like trust, intelligence and social entity. The focus groups will be conducted in a school setting (approx. 90 min.), where two researchers from the consortium will be present. The interviews will be voice-recorded, coded and interpreted. Expected insights: Get a deeper understanding of how the kids perceive different conversational agents with different embodiments, how trustworthy the CAs are for children etc.
- Result – Insights on schoolchildren’s learning strategies with AI-based conversational agents. We will gain a better understanding on how schoolchildren learn with AI-enabled CAs compared with traditional learning methods and what positive and negative consequences this has. To achieve this objective, we will apply experimental methods and carefully manipulate the type of learning assistance and in the case of a CA assistance (in)consistent answers. The “type of assistance” treatment should provide insights into how assisted learning differs between traditional methods and with CAs in terms of, for example learning achievement, knowledge retention, and general mental well-being. The “answer consistency” should provide insights into the persuasiveness of the CAs input. Since the CA input in the inconsistent treatment should be incorrect, the experiment will provide us insights if subjects tend to stick with the CA answer suggesting that a CA is highly persuasive for teenagers or rely on the human input. Consequently, the findings will help us understand if teenagers develop (or not) overtrust and overreliance on CAs. We plan to rely on our in-house developed CA platform in contrast to CAs that are available on the market (e.g., Alexa). Given the high control we plan to have over the CA’s answers, we believe that the insights we will gain are original and novel. Consequently, the findings will provide meaningful insights into the bright and dark consequences of CAs for schoolchildren.
- Result – Dissemination to a wide audience. We will share the insights directly with participating parents and teachers. We will also use the storytelling approach in disseminating them on a dedicated blog. We recognise the need to equip schoolchildren as digital natives and their caregivers with hands-on knowledge in dealing with the rapid advancement of smart technologies, especially those that appear to be social and human-like in either appearance or intelligence. We will synthesise the hands-on knowledge from project results 1, 2, and 3, reinterpreting the insights to be accessible to schoolchildren, their parents, and their teachers. The hands-on knowledge will be published on our project blog using vocabulary, writing style, and formats that are appealing to our target groups. We will promote our project blog using social media and established channels using the networks of our universities. We will dedicate a storytelling section to our project blog. The storytelling section will consist of several short stories about the interaction between two children (a girl and a boy) with their conversational agent friends. The tales will be told from the children’s first-person perspective. Each tale highlights a specific episode within a period of no longer than a day (e.g., doing homework, gardening, choosing a birthday present). We choose such a brief duration so that we are able to elaborate on the behaviours, thoughts, and emotions of the characters.