Intelligenza Artificiale

Reasoning about time for calendar management can be complex, especially when handling multi-user events, because a possibly large number of temporal constraints might have to be analyzed to find solutions which satisfy the commitments of the involved people. Artificial Intelligence temporal reasoning techniques are the key to address this issue but, due to the personal nature of calendar management, they have to be applied in a model that grants the user control in the scheduling decisions to be made. This paper explores the use of Intelligent User Interfaces to mitigate the problem by supporting the management of a dialog between user and automated calendar manager during the scheduling process. The idea is that of enabling the user to steer the calendar manager in the selection of the scheduling solutions to be applied when a temporal conflict occurs. In this way, the user benefits from automated reasoning support but controls the system inferences on the basis of her/his own preferences. The paper also describes the results of a user test we carried out to assess the suitability of the proposed model.

Novel research works in recommender systems have illustrated the benefits of exploiting contextual information, such as the time and location of a suggested place of interest, in order to better predict the user ratings and produce more relevant recommendations. But, when deploying a context-aware system one must put in place techniques for operating in the cold-start phase, i.e., when no or few ratings are available for the items listed in the system catalogue and it is therefore hard to predict the missing ratings and compose relevant recommendations. This problem has not been directly tackled in previous research. Hence, in order to address it, we have designed and implemented several novel algorithmic components and interface elements in a fully operational points of interest (POI) mobile recommender system (STS). In particular, in this article we illustrate the benefits brought by using the user personality and active learning techniques. We have developed two extended versions of the matrix factorisation algorithm to identify what items the users could and should rate and to compose personalised recommendations. While context-aware recommender systems have been mostly evaluated offline, a testing scenario that suffers from many limitations, in our analysis we evaluate the proposed system in live user studies where the graphical user interface and the full interaction design play a major role. We have measured the system effectiveness in terms of several metrics such as: the quality and quantity of acquired ratings-in-context, the recommendation accuracy (MAE), the system precision, the perceived recommendation quality, the user choice satisfaction, and the system usability. The obtained results confirm that the proposed techniques can effectively overcome the identified cold-start problem.

Current users of digital devices have to face the management of a huge amount of heterogeneous digital resources, the switch between activity contexts, and the interaction with many different applications and services. This situation leads to a very fragmented interaction experience, which poses a great cognitive overload for the users and risks to cause lack of efficiency and loss of information. Starting from the limitations of both traditional mechanisms for Personal Information Management (based on the notions of files and hierarchical folders) and new proposals (tagging and folksonomies), in this paper we present Semantic T++, a system supporting users in collaboratively handling digital resources, based on the notion of “tables” (thematic Web-based collaborative workspaces), populated by “objects” (shared digital resources). Semantic T++ exploits a formal semantic representation of such objects to support users in organizing, selecting and using them. Its core is represented by an ontology which models table objects as “information elements” having properties and relations mainly (but not only) related to their content. Reasoning techniques can be applied to infer knowledge useful to provide users with a flexible access to table objects, based on different criteria, which can be defined and combined by the user on the basis of her needs. In order to evaluate our model, we demonstrated its technical feasibility by developing a proof-of-concept prototype, and we showed its advantages in the access to personal and shared resources by discussing the results of a user test.

Cognitive distraction during the driving task might cause impairment of detection performance and of the recognition and/or response selection, increasing the risk of road crashes. In order to avoid or mitigate the negative effects related to cognitive distraction, this paper describes the development and testing of a Cooperative Lane Change Assistant (C-LCA) system: it takes into account the real-time driver's cognitive state by means of a cognitive distraction classifier expressly designed and it implements road cooperation between the vehicles thanks to a cooperative driver model. Three different test sessions were conducted on a static driving simulator and, in each test session, the participants carried out several analogous runs of a reference protocol test, derived from the Lane Change Task. Using the data collected during the first test session, the cognitive distraction classifier was developed using Machine Learning techniques. In the remaining two sessions, a specific C-LCA HMI prototype with visual and acoustic interfaces has been evaluated. The results show that the C-LCA reduced the workload during the lane change manoeuvres compared both with the baseline and with the assistance of a non-cooperative warning system. As well, the users expressed satisfaction about the Visual Interface and Acoustic Interface designed for the C-LCA.

Carers - people who provide regular support for a friend or relative who could not manage without them - frequently report high levels of stress. Good emotional support could help relieve this stress. This study uses seven scenarios that depict different types of stress and acquires emotional support messages for them. We then categorize and evaluate the emotional support for different types of stress. We found that telling the carer they are appreciated and offering support are the best types of emotional support. Additionally, we found that how well a supporter sympathises with a situation affects the type of support they consider suitable. We describe and evaluate an algorithm that selects different categories of support to be used by an intelligent virtual agent to provide emotional support to carers experiencing different types of stress.

This paper discusses an approach to monitoring the level of engagement of video game players based on the theory of flow in the gaming experience. Starting from the flow framework, we developed a non-obtrusive system that estimates the player's state of engagement by analysing non-verbal behavioral cues that are easily detected with simple hardware, such as a webcam and a traditional keyboard and mouse setup. We present the design and the results of an empirical study aimed at gathering data and model the player's engagement. Facial expressions, head movements, keyboard and mouse activities were recorded while participants played a first-person shooter video game. We used an adapted version of the Experience Sampling Methodology to gather the ground truth and trained a Support Vector Machine classifier that recognizes the affective states, reaching an accuracy of 73%. The results showed that the level of engagement is reasonably predicted by considering the head movements and facial expressions only. The findings could aid in developing digital games able to use the information about the player's affective state to adapt their content and support the game experience.