Dear Apple Users, please be aware that there is an iCal Time Zone problem, follow this link.
- Diese Veranstaltung hat bereits stattgefunden.
MCI-SE08: Recommender Systems & Machine Learning
September 8 @ 10:45 - 12:00
Link to the online meeting (Zoom/MS Teams/etc.)
10:45 – 11:00 Uhr
Modeling User Interaction at the Convergence of Filtering Mechanisms, Recommender Algorithms and Advisory Components
Universität Duisburg-Essen, Germany
A variety of methods is used nowadays to reduce the complexity of product search on e-commerce platforms, allowing users, for example, to specify exactly the features a product should have, but also, just to follow the recommendations automatically generated by the system. While such decision aids are popular with system providers, research to date has mostly focused on individual methods rather than their combination. To close this gap, we propose to support users in choosing the right method for the current situation. As a first step, we report in this paper a user study with a fictitious online shop in which users were able to flexibly use filter mechanisms, rely on recommendations, or follow the guidance of a dialog-based product advisor. We show that from the analysis of the interaction behavior, a model can be derived that allows predicting which of these decision aids is most useful depending on the user’s situation, and how this is affected by demographics and personality.
11:00 – 11:15 Uhr
Design of a Knowledge-based Recommender System for Recipes from an End-User Perspective
Universität Siegen, Germany
Nowadays, recommender systems are a fundamental part of several online portals and services. However, most of these systems rely on collective user data and ratings or a preselection of parameters to derive appropriate recommendations. Within this paper, we examine recommendations without previous user data. Therefore, we design, implement, and evaluate a knowledge-based recommender system by turning to recipe recommendations that offer alternatives for favorite recipes. We introduce and compare three versions of a given algorithm. Our evaluation shows that the knowledge-based approach may serve as a good start for deriving appropriate recommendations without prior user data. Moreover, the asymmetric similarity is more user-specific but carries a higher risk of a filter bubble.
11:15 – 11:30 Uhr
“I Never Thought about Securing my Machine Learning Systems”: A Study of Security and Privacy Awareness of Machine Learning Practitioners
1Fraunhofer AISEC, Germany; 2Fraunhofer SIT, Germany; 3Freie Universität Berlin, Germany
Machine learning (ML) models have become increasingly important components of many software systems. Therefore, ensuring privacy and security is a crucial task. Current research mainly focuses on the development of security and privacy methods. However, ML practitioners, as the individuals in charge of translating the theory into practical applications, have not yet received much attention. In this paper, the security and privacy awareness and practices of ML practitioners are studied through an online survey with the aim of (1) gaining insight into the current state of awareness, (2) identifying influencing factors, and (3) exploring the actual use of existing methods and tools. The results indicate a relatively low general privacy and security awareness among the ML practitioners surveyed. In addition, they are less familiar with ML privacy protection methods than with general security methods or ML-related ones. Moreover, awareness correlates with the years of working with ML but not with the level of academic education or the field of occupation. Finally, the practitioners in this study seem to experience uncertainties in implementing legal frameworks, such as the European General Data Protection Regulation, into their ML workflows.
11:30 – 11:45 Uhr
The Impact of Multiple Parallel Phrase Suggestions on Email Input and Composition Behaviour of Native and Non-Native English Writers
1University of Bayreuth, Deutschland; 2Ludwig-Maximilians-University Munich
We present an in-depth analysis of the impact of multi-word suggestion choices from a neural language model on user behaviour regarding input and text composition in email writing. Our study for the first time compares different numbers of parallel suggestions, and use by native and non-native English writers, to explore a trade-off of “efficiency vs ideation”, emerging from recent literature. We built a text editor prototype with a neural language model (GPT-2), refined in a prestudy with 30 people. In an online study (N=156), people composed emails in four conditions (0/1/3/6 parallel suggestions). Our results reveal (1) benefits for ideation, and costs for efficiency, when suggesting multiple phrases; (2) that non-native speakers benefit more from more suggestions; and (3) further insights into behaviour patterns. We discuss implications for research, the design of interactive suggestion systems, and the vision of supporting writers with AI instead of replacing them.
11:45 – 11:48 Uhr
Deep learning meets private talk: Can Conversational AIs predict speaker traits by eavesdropping for as short as 30 seconds?
The Hong Kong Polytechnic University, Hong Kong S.A.R. (China)
Conversational AI such as smart speakers placed in home environments can accidentally activate and record people’s talk for a short time. What can such devices learn about people by listening in on ongoing conversations? Taking two commonly used speaker traits as an example, we present the results of an experiment that simulates Conversational AI eavesdropping on ongoing talk using transcriptions of naturalistic conversations in private settings. We show that a currently popular type of deep learning-based system can reliably predict if a speaker is young, old, female or male (age=99\%, gender=82\%) based on what they say in around 30 seconds. Our results exemplify how powerful current big data language models are when it comes to data-driven predictions of personal information based on how people talk, even when listening only for a short time. We conclude the experiment with a critical comment on the increasingly pervasive use of such user modeling technology to compute speaker traits, touching upon some potential ethical concerns, bias, and privacy issues.
11:48 – 11:51 Uhr
Human-machine collaboration on data annotation of images by semi-automatic labeling
1TU Wien, Austria; 2Fraunhofer Institute fuer Kognitive Systeme IKS
Deployment of deep neural network architectures in computer vision applications requires labeled images which human workers create in a manual, cumbersome process of drawing bounding boxes and segmentation masks. In this work, we propose an image labeling companion that supports human workers to label images faster and more efficiently. Our data-pipeline utilizes One-Shot, Few-Shot and pre-trained object detection models to provide bounding box suggestions, thereby reducing the required user interactions during labeling to corrective adjustments. The resulting labels are then used to continuously update the underlying suggestion models. Optionally, we apply a refinement step, where an available bounding box is converted into a finer segmentation mask. We evaluate our approach with a group of participants who label images using our tool – both manually and with the system. In all our experiments, the achieved quality is consistently comparable with manually created labels at factor 2 to 6 faster execution times.
11:51 – 11:54 Uhr
Let’s Chat Internal: User Acceptance of an In-Company Service Desk Chatbot
Due to remote working models, which were strengthened by the COVID-19 pandemic, it became highly relevant to digitalize on-site service desk consultations. Virtual enterprise assistants which simplify the interaction with in-company services provide similar advantages to their popular peers in customer service: high availability and customer satisfaction, and low response time and costs. To find out how their usage affects user acceptance we developed an early high-fidelity prototype for an in-company UX service desk chatbot and evaluated it in an online survey (N=53) and user study (N=14) with employees from the industry using the Technology Acceptance Model. The results show that the prototype’s acceptance does not differ from the classical service and prior experiences with textual chatbots in a private usage context but was rated higher than a private sample bot.
11:54 – 11:57 Uhr
Too Bureaucratic to Flexibly Learn about AI? The Human-Centered Development of a MOOC on Artificial Intelligence in and for Public Administration
The public sector holds enormous potential for the use of artificial intelligence, which is also recognized and supported by the government. To realize this potential, however, it is imperative that civil servants have the necessary knowledge to recognise and optimally exploit the application and utilisation possibilities of AI. Massive open online courses (MOOC) are a promising way to help civil servants gain that required knowledge. But how can such a course be designed to become accepted by this target group? In this paper, we present a human-centered development approach to develop a MOOC about AI for civil servants. Using an analysis of the target audience’s mental models, knowledge needs, and attitudes, we iteratively developed short learning units that ground the abstract AI topics in concrete case scenarios taken from the public sector. First results of an expert evaluation (expertise in adult education, public sector, and AI) look promising and further evaluations with the target group are planned.
11:57 – 12:00 Uhr
Wisdom of the IoT Crowd: Envisioning a Smart Home-based Nutritional Intake Monitoring System
Obesity and overweight are two factors linked to various health problems that lead to death in the long run. Technological advancements have granted the chance to create smart interventions. These interventions could be operated by the Internet of Things (IoT) that connects different smart home and wearable devices, providing a large pool of data. In this work, we use IoT with different technologies to present an exemplary nutrition monitoring intake system. This system integrates the input from various devices to understand the users’ behavior better and provide recommendations accordingly. Furthermore, we report on a preliminary evaluation through semi-structured interviews with six participants. Their feedback highlights the system’s opportunities and challenges.