This paper introduces CWGPT, a ChatGPT-4-based tool designed for Cognitive Walkthrough (CW) inspired evaluations of web interfaces. The primary goal is to assist users, particularly students and inexperienced designers, in evaluating web interfaces. Our tool, operating as a conversational agent, provides detailed evaluations of a user-specified task by intelligently guessing the subtasks and actions required to accomplish them, answering the standard CW questions, and providing helpful feedback and practical suggestions to improve the usability of the analyzed interface. For our study, we selected a group of web applications designed by students from a Web and Software Architecture course. We compare the outcome of the CWs we executed on ten web apps against the corresponding CWGPT analyses. We then describe the study we conducted involving five author-students to assess the tool’s efficacy in helping them recognize and solve usability issues. In addition to introducing a novel adaptation of ChatGPT, the outcomes of the described experience underscore the promising potential of AI in usability evaluations.
In this work, we analyze the different contexts in which one chooses to integrate artificial intelligence into an interface and the implications of this choice in managing user interaction. While AI in systems can provide significant benefits, it is not infallible and can make errors that seriously affect users. We aim to understand how to design more robust human-AI systems so that these initial AI errors do not lead to more catastrophic failures. To prevent failures, it is essential to detect errors as early as possible and have clear mechanisms to repair them. However, detecting errors in AI systems can be challenging. Therefore, we examine various approaches to error detection and repair, including post-hoc estimation, the use of traces and ambiguity, and multiple sensor layers.
Cruising-for-parking in an urban area is a time-consuming and frustrating activity. We present four machine learning-based models to predict the parking availability of street segments in an urban area on a three-level scale, which navigator and smart-parking apps can exploit to ease and reduce the cruising phase. The models were trained with data generated by a cruising-for-parking simulator that we developed, replicating four parking behavior types (workers, residents, buyers, and visitors). The generated data is comparable to that collectible with smartphones’ sensors. We simulated 40 users moving for 200 weeks in the city area of San Giovanni in Rome. We collected information about users’ parking, unparking, and cruising actions over considered road segments at different time slots. Once a significant amount of trips were collected, we extracted ten features for each road segment at a given time slot. With the obtained dataset, which contained 761 samples, we trained and compared four supervised machine learning models that receive the history of a segment and, in return, classify the Parking Availability Level of the segment as Green, Yellow or Red. The four models were further evaluated in a different city area, San Lorenzo, and obtained very accurate results. We can predict parking availability with an accuracy above 97% for all the street segments where we collected 30 or more user actions, confirming the robustness of the simulator in generating synthetic cruising-for-parking data and the suitability of designing a Parking Availability Classifier (PAC) based on data collectible by smartphones.
Interacting with a smart parking system to find a parking spot might be tedious and unsafe if performed while driving. We present a system based on a Boosted Tree classifier that runs on the smartphone and automatically detects when the driver is cruising for parking. The system does not require direct intervention from the driver and is based on the analysis of context data. The classifier was trained and tested on real data (615 car trips) collected by 9 test users. With this research, we contribute (i) by providing a literature review on cruising detection, (ii) by proposing an approach to model cruising behavior, and (iii) by describing the design, training, and testing of the classifier and discussing its results. In the long term, our work aims to improve user experience and safety in car-related contexts by relying on human-centered features that implicitly understand users’ behavior and anticipate their needs.