Perceived conversational ability of task-based chatbots – Which conversational elements influence the success of text-based dialogues?

The use of text-based chatbots offering individual support to customers has increased steadily in recent years. However, thus far, research has focused on comparing text-based chatbots with either each other or with humans, whilst the investigation of task-based dialogues has been scarce. This paper aims to identify the characteristics of dialogues – that is, conversational elements – that lead to a successful task-based conversation. For this purpose, the chatbot, KIM, by MAGGI Kochstudio was used. It was designed to help customers find a recipe tailored to their individual needs. In order to investigate which conversational elements contribute to successful communication between the user and the chatbot KIM, a usability study collecting 123 unstructured dialogues and a scenario-based experiment using four dialogues with 627 respondents was conducted. The quantitative analysis demonstrates that task completion is characterized by a higher perception of the chatbot’s conversational ability and user satisfaction. The chatbot should propose correct recipe suggestions following a short dialogue, without the user needing to provide too much input. Based on these findings, we recommended equipping the skillset of task-based chatbots with elements that will complement their assistive qualities – for example, improved use of standard phrases, and reactions to similar domains and non-requests. Gender-specific differences in task completion should be considered.     

A systematic literature review of K-12 environmental Citizen Science (CS) initiatives: Unveiling the CS pedagogical and participatory aspects contributing to students’ environmental citizenship

Environmental Citizen Science (CS) initiatives have been largely embraced in K-12 education, as they are often hypothesized to improve students' knowledge, skills, attitudes, and behaviours to act as “environmental citizens” according to the notion of Environmental Citizenship (EC). However, the potential of environmental CS initiatives to promote Education for Environmental Citizenship (EEC) has not been systematically explored. At the same time, environmental CS initiatives for educational purposes are highly heterogenous and learning is enacted in diverse ways, according to the participatory and the pedagogical components underpinning each initiative. To address the complexity of the field, this review study adopts the PRISMA methodology to synthesize thirty-four empirical studies (n = 34) retrieved from a systematic review of the literature covering the last two decades (2000–2020). The reviewed environmental CS initiatives were subjected to a content analysis to identify their impact on students' EC (e.g., EC competences, actions, outcomes), as well as to unveil the CS initiatives' constitutional components in terms of (a) Participation (e.g., types of students' contributions, level of data collection, frequency of students' participation, modes of student engagement, forms of students’ involvement), and (b) Pedagogy (e.g., learning goals, educational contexts, learning mechanisms, EEC pedagogy). Our analysis shed light to the three territories (Participation, Pedagogy, Environmental Citizenship) underpinning the reviewed CS initiatives as well as to their interrelations. We reflect on these findings, and we provide directions for future research to guide the development of more successful environmental CS initiatives in K-12 education, serving as a vehicle for EC.     

Fair and Explainable Depression Detection in Social Media

Detection at an early stage is vital for the diagnosis of the majority of critical illnesses and is the same for identifying people suffering from depression. Nowadays, a number of researches have been done successfully to identify depressed persons based on their social media postings. However, an unexpected bias has been observed in these studies, which can be due to various factors like unequal data distribution. In this paper, the imbalance found in terms of participation in the various age groups and demographics is normalized using the one-shot decision approach. Further, we present an ensemble model combining SVM and KNN with the intrinsic explainability in conjunction with noisy label correction approaches, offering an innovative solution to the problem of distinguishing between depression symptoms and suicidal ideas. We achieved a final classification accuracy of 98.05%, with the proposed ensemble model ensuring that the data classification is not biased in any manner.