Conversational Assistants Research Articles

The Effects of Anthropomorphised Virtual Conversational Assistants on Consumer Engagement and Trust During Service Encounters

Drawing on social exchange and anthropomorphism theory, this research examines the role of virtual conversational assistants (VCA) as frontline employees. Specifically, we investigate the effects of AI-derived features, such as anthropomorphism, in building Human-Machine relationships. Drawing on a qualitative interpretivist approach, 31 semi-structured interviews were conducted with global users of Siri, Alexa and Google Assistant. Our findings suggest anthropomorphism is an important factor in understanding the development of trust within Human-Machine interactions. More specifically, the effects of a humanised voice, interactive communication capability and cognitive features evoke a sense of social presence that may positively or negatively impact user trust. We propose that the interplay between a user’s perceptions of the bright and dark sides of interacting with an AI-empowered anthropomorphised machine determines categories of trust and subsequent customer engagement behaviours with this embedded form of organisational frontline.

AI Enabled Chatbot for regulating Mental Health

Abstract: The most important healthcare questions affecting present-day human people are those related to mental health. Young adults and individuals of working age are more inclined to experience these troubles that have a detrimental effect on the person, his or her family, job, society, and economy. Conventional mental health care aids, while very successful, cannot be scaled until they meet the growing demand from affected individuals, as explained in the first two years of the COVID-19 epidemic. To begin with, it is supposed that there are 792 million families globally that have mental health issues and ailments. That equates to 1 in 10 people worldwide. Conversational assistants, or chatbots, are a recent science creativeness that has been efficiently adapted for mental health management to present first-level help for such individuals. Compared to face-to-face consultations, people are more forthcoming in online chats. Traditional methods for sentiment analysis and consoling may not be effective with adolescents because they are reluctant to express their negative emotions in personal interaction. As a result, this project places a firm attention on utilizing virtual programs to recognize depressing or anxious symptoms in individuals and control their stress.

ECAsT: a large dataset for conversational search and an evaluation of metric robustness.

The Text REtrieval Conference Conversational assistance track (CAsT) is an annual conversational passage retrieval challenge to create a large-scale open-domain conversational search benchmarking. However, as of yet, the datasets used are small, with just more than 1,000 turns and 100 conversation topics. In the first part of this research, we address the dataset limitation by building a much larger novel multi-turn conversation dataset for conversation search benchmarking called Expanded-CAsT (ECAsT). ECAsT is built using a multi-stage solution that uses a combination of conversational query reformulation and neural paraphrasing and also includes a new model to create multi-turn paraphrases. The meaning and diversity of paraphrases are evaluated with human and automatic evaluation. Using this methodology, we produce and release to the research community a conversational search dataset that is 665% more extensive in terms of size and language diversity than is available at the time of this study, with more than 9,200 turns. The augmented dataset not only provides more data but also more language diversity to improve conversational search neural model training and testing. In the second part of the research, we use ECAsT to assess the robustness of traditional metrics for conversational evaluation used in CAsT and identify its bias toward language diversity. Results show the benefits of adding language diversity for improving the collection of pooled passages and reducing evaluation bias. We found that introducing language diversity via paraphrases returned up to 24% new passages compared to only 2% using CAsT baseline.

Affective social anthropomorphic intelligent system

Human conversational styles are measured by the sense of humor, personality, and tone of voice. These characteristics have become essential for conversational intelligent virtual assistants. However, most of the state-of-the-art intelligent virtual assistants (IVAs) are failed to interpret the affective semantics of human voices. This research proposes an anthropomorphic intelligent system that can hold a proper human-like conversation with emotion and personality. A voice style transfer method is also proposed to map the attributes of a specific emotion. Initially, the frequency domain data (Mel-Spectrogram) is created by converting the temporal audio wave data, which comprises discrete patterns for audio features such as notes, pitch, rhythm, and melody. A collateral CNN-Transformer-Encoder is used to predict seven different affective states from voice. The voice is also fed parallelly to the deep-speech, an RNN model that generates the text transcription from the spectrogram. Then the transcripted text is transferred to the multi-domain conversation agent using blended skill talk, transformer-based retrieve-and-generate generation strategy, and beam-search decoding, and an appropriate textual response is generated. The system learns an invertible mapping of data to a latent space that can be manipulated and generates a Mel-spectrogram frame based on previous Mel-spectrogram frames to voice synthesize and style transfer. Finally, the waveform is generated using WaveGlow from the spectrogram. The outcomes of the studies we conducted on individual models were auspicious. Furthermore, users who interacted with the system provided positive feedback, demonstrating the system’s effectiveness.

Implementation of a conversational virtual assistant for open government data portal: Effects on citizens

Insufficient support services are one of the barriers for citizen’s open government data (OGD) utilisation. Considering the strengths and effectiveness of virtual assistants (VAs) in providing instant and user-friendly support services, this study focuses primarily on its implementation for OGD portals. A conversational VA prototype is specifically developed. A between-subjects experiment with one factor (VA/text-based support services) was carried out. Results show citizens’ positive attitudes towards VAs providing support services on OGD portals. No significant difference was found for citizens’ acceptance, trustworthiness and rapport for OGD portals with and without a VA. Accuracy rate of completing tasks with VAs is slightly higher. A correlation lies between trustworthiness and rapport with OGD portals. This study benefits the improvement of OGD portals’ online services to enhance citizens’ experiences and provides a research model both for investigating VAs’ effects and for comparing different designs of a same kind of service for an interface.

Strategic Communication and Artificial Intelligence: Reviewing Emerging Innovations and Future Directions

This review examines the adoption of artificial intelligence (AI) as a novel technology for strategic communication. This study includes a survey of pertinent literature that highlights innovative communication linked to AI applications in public relations, marketing and branding, political communication, and customer relations with the goal of enlightening both research and practice. Emerald Insight and Google Scholar, two essential online databases, were used to retrieve and analyse information from peer-reviewed journal articles and chapters (n = 17) published between 2017 and 2022. Based on the main inquiries raised by the research, significant themes were discovered and critically explored. Political communication with AI, AI-focused public relations and marketing approach, and conversational AI assistants were the major themes in the research on strategic communication and AI integration. These studies were primarily the outcome of quantitative and mixed-method techniques used in a western environment. This review satisfies the need for research on AI and strategic communication that makes inferences applicable to the context in Ghana. The study's conclusion outlines a research agenda for the future, emphasising the need for research on particular topics like integration of AI voice assistants in relationship marketing domains, adoption of wearable technology and its relevance to customers' health consciousness as well as privacy concerns of emerging innovation. Future studies should also focus on developing more evidence-based AI suggestions for strategic communication practitioners, as well as a thorough methodological and theoretical discussion of how to further close the gap between research and practice in the African setting.

Alexa doesn't have that many feelings: Children's understanding of AI through interactions with smart speakers in their homes

As voice-based Conversational Assistants (CAs), including Alexa, Siri, Google Home, have become commonly embedded in households, many children now routinely interact with Artificial Intelligence (AI) systems. It is important to research children's experiences with consumer devices which use AI techniques because these shape their understanding of AI and its capabilities. We conducted a mixed-methods study (questionnaires and interviews) with primary-school children aged 6–11 in Scotland to establish children's understanding of how voice-based CAs work, how they perceive their cognitive abilities, agency and other human-like qualities, their awareness and trust of privacy aspects when using CAs and what they perceive as appropriate verbal interactions with CAs. Most children overestimated the CAs' intelligence and were uncertain about the systems' feelings or agency. They also lacked accurate understanding of data privacy and security aspects, and believed it was wrong to be rude to conversational assistants. Exploring children's current understanding of AI-supported technology has educational implications; such findings will enable educators to develop appropriate materials to address the pressing need for AI literacy.

The More Competent, the Better? The Effects of Perceived Competencies on Disclosure Towards Conversational Artificial Intelligence

Conversational AI (e.g., Google Assistant or Amazon Alexa) is present in many people’s everyday life and, at the same time, becomes more and more capable of solving more complex tasks. However, it is unclear how the growing capabilities of conversational AI affect people’s disclosure towards the system as previous research has revealed mixed effects of technology competence. To address this research question, we propose a framework systematically disentangling conversational AI competencies along the lines of the dimensions of human competencies suggested by the action regulation theory. Across two correlational studies and three experiments ( Ntotal = 1453), we investigated how these competencies differentially affect users’ and non-users’ disclosure towards conversational AI. Results indicate that intellectual competencies (e.g., planning actions and anticipating problems) in a conversational AI heighten users’ willingness to disclose and reduce their privacy concerns. In contrast, meta-cognitive heuristics (e.g., deriving universal strategies based on previous interactions) raise privacy concerns for users and, even more so, for non-users but reduce willingness to disclose only for non-users. Thus, the present research suggests that not all competencies of a conversational AI are seen as merely positive, and the proposed differentiation of competencies is informative to explain effects on disclosure.

Conversational AI Assistant

Abstract: A conversational assistant is an intelligent conversational computing system designed to mimic human speech to provide automated online guidance and support. The growing benefits of conversational support have led to widespread adoption across many industries to provide virtual support to their customers. Conversation assistance uses methods and algorithms from his two fields of artificial intelligence: natural language processing and machine learning. However, the application has many challenges and limitations. This research reviews recent advances in conversation support using artificial intelligence and natural language processing. We highlight the main challenges and limitations of the current work and provide recommendations for future research investigations.

Improving ​conversational ​search with ​query ​reformulation ​using ​selective ​contextual ​history

Automated responses to questions for conversational agents, known as conversation passage retrieval, is challenging due to omissions and implied context in user queries. To help address this challenge, queries can be re-written using pre-trained sequence-to-sequence models based on contextual clues from the conversation's history to resolve ambiguities. In this research, we use the TREC conversational assistant (CAsT) 2020 dataset, selecting relevant single sentences from conversation history for query reformulation to improve system effectiveness and efficiency by avoiding topic drift. We propose a practical query selection method that measures clarity score to determine whether to use response sentences during reformulation. We further explore query reformulation as a binary term classification problem and the effects of rank fusion using multiple retrieval models. T5 and BERT retrievals are inventively combined to better represent user information need. Using multi-model fusion, our best system outperforms the best CAsT 2020 run, with an NDCG@3 of 0.537. The implication is that a more selective system that varies the use of responses depending on the query produces a more effective conversational reformulation system. Combining different retrieval results also proved effective in improving system recall.

A Human-Centric Perspective on Model Monitoring

Predictive models are increasingly used to make various consequential decisions in high-stakes domains such as healthcare, finance, and policy. It becomes critical to ensure that these models make accurate predictions, are robust to shifts in the data, do not rely on spurious features, and do not unduly discriminate against minority groups. To this end, several approaches spanning various areas such as explainability, fairness, and robustness have been proposed in recent literature. Such approaches need to be human-centered as they cater to the understanding of the models to their users. However, there is little to no research on understanding the needs and challenges in monitoring deployed machine learning (ML) models from a human-centric perspective. To address this gap, we conducted semi-structured interviews with 13 practitioners who are experienced with deploying ML models and engaging with customers spanning domains such as financial services, healthcare, hiring, online retail, computational advertising, and conversational assistants. We identified various human-centric challenges and requirements for model monitoring in real-world applications. Specifically, we found that relevant stakeholders would want model monitoring systems to provide clear, unambiguous, and easy-to-understand insights that are readily actionable. Furthermore, our study also revealed that stakeholders desire customization of model monitoring systems to cater to domain-specific use cases.

Open-domain conversational search assistants: the Transformer is all you need

Open-domain conversational search assistants: the Transformer is all you need

Increasing Context for Estimating Confidence Scores in Automatic Speech Recognition

Accurate confidence measures for predictions from machine learning techniques play a critical role in the deployment and training of many speech and language processing applications. For example, confidence scores are important when making use of automatically generated transcriptions in training automatic speech recognition (ASR) systems, as well as down-stream applications, such as information retrieval and conversational assistants. Previous work on improving confidence scores for these systems has focused on two main directions: designing features correlated with improved confidence prediction; and employing sequence models to account for the importance of contextual information. Few studies, however, have explored incorporating contextual information more broadly, such as from the future, in addition to the past, or making use of alternative multiple hypotheses in addition to the most likely one. This article introduces two general approaches for encapsulating contextual information from lattices. Experimental results illustrating the importance of increasing contextual information for estimating confidence scores are presented on a range of limited resource languages where word error rates range between 30% and 60%. The results show that the novel approaches provide significant gains in the accuracy of confidence estimation.

Adoption of Conversational Assistants in Banking Based on Extended Technology Acceptance Model: Perceived Risk as a Moderator

Adoption of Conversational Assistants in Banking Based on Extended Technology Acceptance Model: Perceived Risk as a Moderator

I Wouldn’t Search That With My Mobile Phone: Credibility and Trust in OHIRs Among Lower-Income Black Older Adults

Online health information resources (OHIRs) such as conversational assistants and smart devices that provide access to consumer health information in the home are promoted as viable options for older adults to independently manage health. However, there is question as to how well these devices are perceived to meet the needs of marginalized populations such as lower-income Black older adults who often experience lower digital literacy or technology proficiency. We examined the experiences of 34 lower-income Black older adults aged 65-83 from Chicago and Detroit with various OHIRs and explored whether conversational resources were perceived to better support health information seeking compared to traditional online web searching. In a three-phase study, participants tracked their experiences with various OHIRs and documented health-related questions in a health diary. Participants were then interviewed about their diaries in focus groups and semi-structured interviews, followed by a technology critique and co-design session to re-envision a more usable and engaging conversational device. We present preliminary results of the themes that emerged from our analysis: cultural variables in health information seeking practices, perceptions of credibility, likelihood of use, and system accessibility. Participants indicated that their trust of different resources depended on the type of information sought, and that conversational assistants would be a useful resource that require less technology proficiency, even among those with lower e-health literacy. Although our findings indicate that familiarity and trust were salient constructs associated with perceptions of OHIRs, these devices may address digital literacy and technology familiarity with certain design considerations.

Trust Me on This One: Conforming to Conversational Assistants

Conversational artificial agents and artificially intelligent (AI) voice assistants are becoming increasingly popular. Digital virtual assistants such as Siri, or conversational devices such as Amazon Echo or Google Home are permeating everyday life, and are designed to be more and more humanlike in their speech. This study investigates the effect this can have on one’s conformity with an AI assistant. In the 1950s, Solomon Asch’s already demonstrated the power and danger of conformity amongst people. In these classical experiments test persons were asked to answer relatively simple questions, whilst others pretending to be participants tried to convince the test person to give wrong answers. These studies were later replicated with embodied robots, but these physical robots are still rare. In light of our increasing reliance on AI assistants, this study investigates to what extent an individual will conform to a disembodied virtual assistant. We also investigate if there is a difference between a group that interacts with an assistant that communicates through text, one that has a robotic voice and one that has a humanlike voice. The assistant attempts to subtly influence participants’ final responses in a general knowledge quiz, and we measure how often participants change their answer after having been given advice. Results show that participants conformed significantly more often to the assistant with a human voice than the one that communicated through text.

Mitigating Patient and Consumer Safety Risks When Using Conversational Assistants for Medical Information: Exploratory Mixed Methods Experiment.

BackgroundPrior studies have demonstrated the safety risks when patients and consumers use conversational assistants such as Apple’s Siri and Amazon’s Alexa for obtaining medical information.ObjectiveThe aim of this study is to evaluate two approaches to reducing the likelihood that patients or consumers will act on the potentially harmful medical information they receive from conversational assistants.MethodsParticipants were given medical problems to pose to conversational assistants that had been previously demonstrated to result in potentially harmful recommendations. Each conversational assistant’s response was randomly varied to include either a correct or incorrect paraphrase of the query or a disclaimer message—or not—telling the participants that they should not act on the advice without first talking to a physician. The participants were then asked what actions they would take based on their interaction, along with the likelihood of taking the action. The reported actions were recorded and analyzed, and the participants were interviewed at the end of each interaction.ResultsA total of 32 participants completed the study, each interacting with 4 conversational assistants. The participants were on average aged 42.44 (SD 14.08) years, 53% (17/32) were women, and 66% (21/32) were college educated. Those participants who heard a correct paraphrase of their query were significantly more likely to state that they would follow the medical advice provided by the conversational assistant (χ21=3.1; P=.04). Those participants who heard a disclaimer message were significantly more likely to say that they would contact a physician or health professional before acting on the medical advice received (χ21=43.5; P=.001).ConclusionsDesigners of conversational systems should consider incorporating both disclaimers and feedback on query understanding in response to user queries for medical advice. Unconstrained natural language input should not be used in systems designed specifically to provide medical advice.

Considerations on creating conversational agents for multiple environments and users

AbstractAdvances in artificial intelligence algorithms and expansion of straightforward cloud‐based platforms have enabled the adoption of conversational assistants by both, medium and large companies, to facilitate interaction between clients and employees. The interactions are possible through the use of ubiquitous devices (e.g., Amazon Echo, Apple HomePod, Google Nest), virtual assistants (e.g., Apple Siri, Google Assistant, Samsung Bixby, or Microsoft Cortana), chat windows on the corporate website, or social network applications (e.g. Facebook Messenger, Telegram, Slack, WeChat).Creating a useful, personalized conversational agent that is also robust and popular is nonetheless challenging work. It requires picking the right algorithm, framework, and/or communication channel, but perhaps more importantly, consideration of the specific task, user needs, environment, available training data, budget, and a thoughtful design.In this paper, we will consider the elements necessary to create a conversational agent for different types of users, environments, and tasks. The elements will account for the limited amount of data available for specific tasks within a company and for non‐English languages. We are confident that we can provide a useful resource for the new practitioner developing an agent. We can point out novice problems/traps to avoid, create consciousness that the development of the technology is achievable despite comprehensive and significant challenges, and raise awareness about different ethical issues that may be associated with this technology. We have compiled our experience with deploying conversational systems for daily use in multicultural, multilingual, and intergenerational settings. Additionally, we will give insight on how to scale the proposed solutions.

Assessing Top- Preferences

Assessors make preference judgments faster and more consistently than graded judgments. Preference judgments can also recognize distinctions between items that appear equivalent under graded judgments. Unfortunately, preference judgments can require more than linear effort to fully order a pool of items, and evaluation measures for preference judgments are not as well established as those for graded judgments, such as NDCG. In this article, we explore the assessment process for partial preference judgments, with the aim of identifying and ordering the top items in the pool, rather than fully ordering the entire pool. To measure the performance of a ranker, we compare its output to this preferred ordering by applying a rank similarity measure. We demonstrate the practical feasibility of this approach by crowdsourcing partial preferences for the TREC 2019 Conversational Assistance Track, replacing NDCG with a new measure named compatibility . This new measure has its most striking impact when comparing modern neural rankers, where it is able to recognize significant improvements in quality that would otherwise be missed by NDCG.

Intrinsic motivation in virtual assistant interaction for fostering spontaneous interactions.

With the growing utility of today's conversational virtual assistants, the importance of user motivation in human-artificial intelligence interactions is becoming more obvious. However, previous studies in this and related fields, such as human-computer interaction, scarcely discussed intrinsic motivation (the motivation to interact with the assistants for fun). Previous studies either treated motivation as an inseparable concept or focused on non-intrinsic motivation (the motivation to interact with the assistant for utilitarian purposes). The current study aims to cover intrinsic motivation by taking an affective engineering approach. A novel motivation model is proposed, in which intrinsic motivation is affected by two factors that derive from user interactions with virtual assistants: expectation of capability and uncertainty. Experiments in which these two factors are manipulated by making participants believe they are interacting with the smart speaker "Amazon Echo" are conducted. Intrinsic motivation is measured both by using questionnaires and by covertly monitoring a five-minute free-choice period in the experimenter's absence, during which the participants could decide for themselves whether to interact with the virtual assistants. Results of the first experiment showed that high expectation engenders more intrinsically motivated interaction compared with low expectation. However, the results did not support our hypothesis that expectation and uncertainty have an interaction effect on intrinsic motivation. We then revised our hypothetical model of action selection accordingly and conducted a verification experiment of the effects of uncertainty. Results of the verification experiment showed that reducing uncertainty encourages more interactions and causes the motivation behind these interactions to shift from non-intrinsic to intrinsic.