The Digital Communication Methods Lab, an initiative of the RPA Communication, is supporting a series of innovative research projects focused on substantive and methodological development for communication science research. The projects for 2018, done in partnership with the Amsterdam School of Communication Research (ASCoR) members, are distributed across the several research topics from the lab:
- Augmented and Virtual Reality Applications
- Automated Content Analysis
- Conversational Agents and AI
- Media Exposure and Mobile Communication
- Social Robots
Augmented and Virtual Reality Applications
Bridges and Roller Coasters Revisited: Testing Excitation Transfer in Virtual Reality
dr. Jeroen Lemmens, dr. Sindy Sumter, dr. Susanne Baumgartner, Zeph van Berlo, MSc
Excitation transfer theory, as first described by Zillmann (1971), posits that residual excitement from an arousing stimulus may serve to intensify a later emotional state. The empirical basis for this theory can be traced back to a classic ‘suspension bridge’ field experiment conducted by Dutton and Aron (1974), and more recently a ‘roller coaster’ experiment by Meston and Frohlich (2003). Both their findings can be interpreted within the framework of excitation transfer theory in that residual arousal intensified participants’ subsequent experience of interpersonal attraction. Despite being considered by many as classic examples of excitation transfer, neither study has meticulously tested the fundamental assumption of physiological excitement. The current research aims to address these issues and further improve the internal validity by moving these classic field experiments into a controlled environment, and testing the assumptions in a virtual reality lab experiment. The aim of our study is twofold: (1) replicating these two classic experiments on excitation transfer using physiological measures of arousal to determine its exact role within misattribution of emotional valence; and (2) broadening the methodological scope of excitation transfer by applying it within a virtual reality setting and by expanding the range of relevant outcome measures to evaluations of commercials. This work will offer crucial implications for one of the most influential theoretical assumptions in communication science, and moreover, this work provides a glimpse into the possibilities of manipulating both interpersonal and brand attraction within the emerging research field of social VR (e.g., Facebook Spaces, AltSpace).
Social Processes in VR: Comparing the Effects of Virtual versus Real-Life Eating Companions on Healthy and Unhealthy Food Intake
dr. Saar Mollen, dr. Nynke van der Laan, and dr. Sindy Sumter
The recent release of affordable head-mounted displays, and the development of multi-user virtual worlds, foretell a future where part of our social interactions will be taking place in Virtual Reality (VR) environments. Social VR is still in a very early stage and little is known about the nature of social processes in VR and whether social factors in VR influence behavior similarly as in real life. For the proposed project, we will focus on eating behavior, which is a health behavior particularly sensitive to social influences. We will investigate whether social models in VR, have similar effects on eating behavior, as real-life models, and whether similar processes (e.g., imitation, social norms) underlie their influence.
To this end we will conduct an experiment in which participants are either exposed to a real-life or virtual eating partner and monitor the extent to which they imitate the other person’s eating behavior. VR allows researchers to measure social processes in a tightly controlled yet realistic situation. If the proposed research confirms that similar processes underlie food intake in VR as in non-mediated settings this means that researchers can relatively easily and cheaply assess and manipulate social influences on health behaviors. The current study therefore not only adds to theory on media and eating behavior and social influences and eating behavior, it also advances methodology used in social science.
Testing Package Complexity in an Innovative 3D Virtual Supermarket Environment
prof. Edith Smit, dr. Nynke van der Laan, and dr. Corine Meppelink
Although the supermarket is a very familiar environment for most people, it can also be a complex one for some. Especially consumers with a specific health goal face challenges in choosing between products with packages full of claims. The aim of the project is to investigate the effect of packaging complexity on purchasing nutrient-marketed foods and the moderating effect of health-literacy in a truly immersive 3D virtual reality supermarket.
The study has a 2 (packaging complexity: high versus low) X 2 (health literacy: high versus low) between-subjects design. Participants are instructed to shop with a specific health goal (e.g., low in saturated fat) and presented in VR with the shelf of a specific food category (e.g., dairy). Outcome measures are whether they picked up the product in line with the diet, how long they kept the product in their hands, and which products they purchased.
Automated Content Analysis
Building a Hype-Detector
dr. Iina Hellsten and dr. Damian Trilling
Digital media enables the monitoring of large streams of messages posted in the various traditional, social and organizational media. However, extracting meaningful information from the data streams is a challenge. We develop an (semi-)automated tool for detecting spiking words and phrases to identify emerging, potential hypes, i.e. simultaneous co-attention peak in several communication networks (e.g. traditional news and online discussions). We distinguish between two types emerging hypes: First, spike of a new word/phrase such as in the case of ‘fipronil’, related to the egg crisis (in 2017). Second, spikes related to existing words/phrases, such as ‘fake news’ (since 2016). We expect sudden changes in the meanings of the words/phrase as an antecedent of the attention peaks. We use natural language processing and machine learning for calculating the observed/expected values for the words, such as tf-idf scores or log-likelihood scores, and the occurrence of specific words signaling hypes.
Keeping Score: Comparing Inductive and Deductive Approaches to Study Dynamic Issue Agendas
dr. Anne Kroon, dr. Toni van der Meer and prof. Rens Vliegenthart
In an area of evolving online politics and digital media, studying the dynamics of issue agendas has become increasingly complex. The application of computational methods helps communication researchers to better handle the sheer volume of this data. Within the increasing variety of methods, the contribution of this project is twofold. First, this project aims to map the specific benefits and drawbacks associated with different techniques in studying issues agendas. Specifically, results yielded by inductive (i.e., automated text categorization) and deductive (i.e., text analysis based on pre-defined categories) methods will be compared. The second objective is to optimize automated over time analyses. Especially the low-cost inductive approaches have proven to be a successful means of identifying topics in large sets of text but less successful in documenting dynamics and trends in issue agendas for longer periods. To do so, this project will rely on data from political, organizational, and media agendas.
Conversational Agents and AI
Conversational Agents in Public Health: Causes, Content, and Contingencies of Chatbot Usage for STD Prevention
dr. Gert-Jan de Bruijn, dr. Catherine Bolman (Open University Netherlands), and Erwin Fisser (Soa Aids Netherlands)
Within the Netherlands, Soa Aids Netherlands is tasked with the application of e-health preventive solutions for sexually transmitted diseases (STD) and recently started employing chatbots to provide those e-health solutions. Through chatbots, face-to-face interpersonal communication is mimicked in human-computer interactions based on natural language interpretations determined by a conversational agent computer program, giving the user the ability to ask questions and receive answers in natural language. Chatbots can also be accessed 24/7, thereby unburden human public health care professionals. This proposal is designed to (1) predict public health chatbot usage (2) apply text mining and topical modeling to classify content of chatbot interactions, and (3) understand the contingencies of those interactions. The key deliverables of this proposal will be (I) an algorithm for sentiment analysis in STD chatbot conversations and (II) topical parameter estimator for STD chatbots. Both will be made available for use in other public health domains.
Development and Pilot-Evaluation of an Interactive, Computer-Simulated
Virtual Patient-based eLearning to Train Clinicians in Communication Skills
prof. Julia C. M. van Weert, prof. Ellen M. A. Smets (AMC-UvA), dr. Gert-Jan de Bruijn, and dr. ir. Willem-Paul Brinkman (TU Delft)
Although Shared decision making (SDM) is endorsed for many health care situations where decisions have to be made, effective use of the principles of SDM is not yet widespread. The use of computer-simulated virtual patients (VP) allows clinicians to train their skills in a highly interactive and immersive way at a chosen time and location. To capitalize on this promising technology, this project will evaluate an interactive VP-based learning module aiming to enhance clinicians’ SDM skills. An initial version of a computer based easy accessible VP for the training of SDM skills has already been created at the TU Delft in a collaboration with the departments of Communication Science (UvA), Medical Psychology (AMC) and Medical Decision Making (LUMC). In the proposed project the prototype of the VP will be evaluated on its feasibility as a first step in a user-centered development process among medical students during and after three sessions with the VP.
Going Beyond One-shot Experiments: Chatbots as Regular and Personalized Interaction Partners
dr. Theo Araujo and dr. Nadine Bol
Conversational agents in the form of chatbots available on social media and messaging apps, or virtual assistants in the phone or in the home are becoming increasingly prevalent in our communication environment. The effects of ongoing interactions with these agents on individuals’ cognitions, attitudes, emotions, and behaviour, however, have yet to be fully explored. This project takes a step in this direction by investigating how repeated and increasingly personalized interactions with a chatbot influence, over time, trust in the agent, recommendation adherence, perceived information credibility and self-disclosure levels. Moreover, from a methodological perspective, this project explores the effectiveness of current bot technology for data collection in an ongoing and longitudinal manner, and aims at building expertise for experiments going beyond one-shot interactions – investigating bots as regular interaction partners.
Media Exposure and Mobile Communication
Theorizing Personalization versus Customization Effects in Mobile Communication Technologies Using Behavioral Tracking Data
Minh Hao Nguyen, MSc, dr. Nadine Bol, and dr. Annemiek Linn
Mobile technology has become one of the most defining communication technologies of our time. Mobile devices allow for tremendous amounts of personal data to be tracked, which can serve as input to deliver personalized content to users (i.e. system-initiated personalization; SIP). Mobile technology also perfectly lends itself for user-customization of in-app features, enabling users to create personalized information environments (i.e., user-initiated customization; UIC). Current literature on SIP and UIC effects have studied these tailoring strategies separately and operationalized these concepts inconsistently. This makes it difficult to compare effects and understand which mechanisms are triggered by SIP and UIC, and under which conditions. This project takes on an innovative approach by combining tracking data with self-report measures to explain what makes SIP and UIC in mobile health and news applications effective and for whom. A new avenue of exploration will be how to use the tracking data to explain these effects.
Social Robots
Does Social Presence Affect Answers to Sensitive Questions? An Experimental Comparison of Face-to-face, Telepresence Robot, and Skype-based Survey Modes
dr. Alex Barco Martelo, dr. Rinaldo Kühne, and prof. Jochen Peter
The validity of answers to sensitive questions in questionnaires has been a long-standing problem in survey research. The proposed project focuses on a questionnaire administered by interviewers who vary in their social presence. The key question of the proposed project is whether, in interviewer-administered questionnaires, the validity of respondents’ answers to sensitive question will decrease systematically as the social presence of the interviewer increases(skype; telepresence robot; face to face). We expect that answers to sensitive questions are most biased in the face-to-face condition, followed by the telepresence robot and the Skype condition. To address our research question, we will perform a one-factorial experiment with three experimental conditions: face-to-face interviewing, telepresence robot-based interviewing, and Skype-based interviewing. Respondents in all conditions will be asked a number of questions that vary in their sensitivity.