Personality traits such as Need for Cognition, Locus of Control, Mindset and Self-eficacy could impact the perception, acceptance and appreciation of recommendations provided to support configuration tasks in the End User Development (EUD) context. In this paper we describe the user model services we have developed to measure such traits. These services can be accessed by users through a simple web interface and can be queried by EUD applications by means of REST API.
While smart environments can be understood as ecosystems of interacting objects, such as
sensors, devices, appliances and embedded systems [
Recommender systems can support users’ decision making processes [
Starting from these observations, in our past work carried out in the context of the EMPATHY
project1 we experimented with personalized recommendations as a way to help users make
configuration choices [
Nevertheless, while recommendations can be broadly considered useful for any end-user,
their actual perception, acceptance and appreciation can depend on individual features. For
example, [
Following on from these results, we put forward that incorporating information on such personality traits in user models can help generate more efective recommendations for endusers. For example, the quantity of recommendations as well as the presence and level of detail of explanations could be tailored according to user personality. As part of our research within the EMPATHY project, we are carrying out a series of experiments aimed at determining the impact of personality traits on recommendations in the specific context of a configuration task, an aspect which has not yet been explored in relevant literature. In addition, we have developed a few web services which allow to measure diferent personality traits, making use of standard scales, and which are publicly made available to other researchers.
In this paper, after having discussed related work (Section n2), we present our web services (Section 3), which are part of a platform we built to experiment with recommendations (Section 4). Conclusions (Section 5) complete the paper.
Personality in recommender systems. Personality explains individual diferences in emotional, interpersonal, experiential, attitudinal and motivational styles [16]. Personality-related information has been used in recommender systems to help overcome various issues, such as the cold start problem [17] or satisfaction with the suggested options in a group recommendation scenario [18]. While most work has referred to the Big-Five model [16], which provides an overall picture of personality in terms of five broad dimensions, some studies have focused on specific traits. For example, Mindset was found to influence users’ evaluation of recommendations at various levels: perceived efectiveness [ 19], likelihood to be influenced [20] and acceptance [14]. Need for Cognition has an impact on user satisfaction with explanations [12] and willingness to rely on recommendations [21]. Locus of Control has an efect on users’ tendency to trust recommendations [13], while, to the best of the authors’ knowledge, Self-eficacy was not studied in connection with recommender systems, but significantly afects users’ acceptance of diferent technologies [15, 22].
Recommendation in the EUD domain Among the earliest examples of recommendations in
the EUD domain are systems such EAGER, Dynamic Macro, and APE [23, 24, 25], which adopt
a programming by example (PBE) approach to learn how to complete users’ tasks based on
the observation of their behaviour and therefore recommended automation directly within the
user’s workflow. In a similar vein, Lumière [ 26] and WARP [27] used probabilistic models for
activity recognition to ofer context-dependent assistance. While Lumière provided suggestions
on a limited number of predefined tasks series of predefined tasks, WARP was also able to
continuously extend its knowledge. Task Assistant also generated recommendations over an
extensible knowledge base, exploiting manual procedures defined through EUD [ 28]. Nevertheless,
Haines et al. [
With the aim of fostering research on the interplay between personality and recommendations in the EUD domain, as well as ofering a public service to researchers and pratictioners who wish to incorporate information on personality in their recommender systems, we have built a mechanism to calculate four diferent, but closely related traits, designing a specific test for each of them: 1. Self eficacy (https://app.empathy.di.unito.it/#/selfEficacy). This construct refers to individuals’ beliefs about their ability to exercise control over their own functioning and activities [29]. To measure it, we used the the 10-item Self-eficacy IPIP scale 2. 2. Need for cognition (https://app.empathy.di.unito.it/#/needForCognition). This construct describes individuals’ tendency to engage in and enjoy thinking and, in general, cognitively demanding activities[30]. To measure it, we used the 10-item IPIP scale3. 3. Locus of control (https://app.empathy.di.unito.it/#/locusOfControl). This construct represents whether individuals believe they have weak (external) or strong (internal) 2https://ipip.ori.org/newCPIKey.htm#Self-Eficacy) 3https://ipip.ori.org/newSingleConstructsKey.htm#Need-for-Cognition control over the events that afect their lives [ 31]. To measure it, we used the 5-item IPIP rational scale4 4. Mindset (https://app.empathy.di.unito.it/#/mindset). This construct refers to implicit theories which create a sort of framework and then stimulate judgments and reactions which are coherent with that framework. To measure it, we formulated a set of questions inspired by the Growth Mindset Scale by Dweck et al. [32], which investigates the beliefs in fixed versus malleable human attributes. Difernetly from the original scale, we focused on problem solving instead of intelligence.
Each test is designed to provide a mechanism to profile their users to external applications. On the one hand, users can perform one or multiple tests, according to the application needs, accessing the aforelisted web pages. On the other hand, the prototype also exposes two REST API which provide the test results to querying applications in order to promote interoperable user modeling exchanges [33], Berkovsky et al. [34]. Results for a specific user can be retrieved by providing their username. Thresholds (“low”, “medium”, “high”) for the four traits were calculated based on the dataset collected in a previous experiment [35].
In particular, one API provides an overview of the personality traits for the selected user: https: //app.empathy.di.unito.it/api/empathy/userPsychometrics/[username]. A response example follows: The other API provides more details on the traits by returning numerical values: https://app.empathy.di.unito.it/api/empathy/userPsychometricsValues/[username]. A response example follows: "locusOfControl": "external", "needForCognition": "low", "selfEfficacy": "low", "username": "Donald.Duck" "locusOfControl": 1.8, "rangeLocusOfControl": "2.885 - 3.615", "needForCognition": 3, "rangeNeedForCognition": "3.46 - 3.98", "selfEfficacy": 3, "rangeSelfEfficacy": "3.52 - 3.96", "username": "Donald.Duck" { } { }
In the context of the EMPATHY project, we have built a prototype platform5 to experiment with recommendations and connected services. Apart from the psychometric user model services described in Section 3, the platform hosts: • A user modeling component (under development) which includes individual characteristics and psychological traits as well as owned smart objects, user preferences for smart objects and smart home goals (such as energy saving, safety, comfort, etc.). • A testbed for similarity algorithms, where the researcher can choose both the preferred measure (Jaccard, Pearson, Cosine, Simple matching) and the user model features to include. • A series of recommendation services which suggest smart objects (either a trigger or an action object) to couple with an input object chosen by the end-user. A web interface where these services can be accessed is also available: more specifically, one page suggests action object categories, given a trigger object category, while the other one suggests the trigger object categories, given an action object category. Example rules are also suggested for each trigger-action association. Knowledge on suitable trigger-action associations was derived by applying the association rules technique (Apriori algorithm) on an IFTTT6 rules dataset.
Implementation details. The web interface is developed using Vue.js and Vuetify as a material design framework. The application server is implemented with Spring Boot and exposes a set of REST API. The recommender is a module built in Java which contains all the logic required to provide the diferent recommendation techniques. The data are stored in a mongoDB instance and are organized in diferent collections and databases.
In this paper, we have presented four web services which calculate the personality traits of Need for Cognition, Locus of Control, Mindset and Self-eficacy, which we believe can impact users’ behaviour in recommender systems in the context of configuration tasks. As future work, we are planning to provide guidelines on the use of personality information to tailor recommendations in the EUD domain, based on the results of the experiments we are currently carrying out.
This work is partially supported by the Italian Ministry of University and Research (MIUR) under grant PRIN 2017 “EMPATHY: EMpowering People in deAling with internet of THings ecosYstems”.
6If This Then That (IFTTT) is a private commercial company that runs services that allow a user to program a response to events (https://ifttt.com/). with respect to attachment styles., Educational Sciences: Theory and Practice 11 (2011) 105–113. [12] M. Millecamp, N. N. Htun, C. Conati, K. Verbert, To explain or not to explain: The efects of personal characteristics when explaining music recommendations, in: Proceedings of the 24th International Conference on Intelligent User Interfaces, IUI ’19, Association for Computing Machinery, New York, NY, USA, 2019, p. 397–407. URL: https://doi.org/10.1145/ 3301275.3302313. doi:10.1145/3301275.3302313. [13] N. N. Sharan, D. M. Romano, The efects of personality and locus of control on trust in humans versus artificial intelligence, Heliyon 6 (2020) e04572. URL: https:// www.sciencedirect.com/science/article/pii/S240584402031416X. doi:https://doi.org/ 10.1016/j.heliyon.2020.e04572. [14] M. Aljukhadar, S. Senecal, The efect of consumer-activated mind-set and product involvement on the compliance with recommender system advice, SAGE Open 11 (2021) 21582440211031550. URL: https://doi.org/10.1177/21582440211031550. doi:10.1177/ 21582440211031550. arXiv:https://doi.org/10.1177/21582440211031550. [15] H. A. Alfadda, H. S. Mahdi, Measuring students’ use of zoom application in language course based on the technology acceptance model (tam), Journal of psycholinguistic research 50 (2021) 883–900. [16] R. McCrae, O. P. John, An introduction to the five-factor model and its applications.,
Journal of personality 60 2 (1992) 175–215. [17] R. Hu, P. Pu, Enhancing collaborative filtering systems with personality information, in: Proceedings of the Fifth ACM Conference on Recommender Systems, RecSys ’11, Association for Computing Machinery, New York, NY, USA, 2011, p. 197–204. URL: https: //doi.org/10.1145/2043932.2043969. doi:10.1145/2043932.2043969. [18] A. Delic, J. Neidhardt, T. N. Nguyen, F. Ricci, An observational user study for group recommender systems in the tourism domain, J. Inf. Technol. Tour. 19 (2018) 87–116. URL: https://doi.org/10.1007/s40558-018-0106-y. doi:10.1007/s40558-018-0106-y. [19] N. Reijmer, Improving energy saving decisions by matching recommender type with domain knowledge and mindset, Ph.D. thesis, Eindhoven University of Technology, 2011. [20] D. Y. Kim, H. Im, Good or better; the efect of comparative mindset with recommended products on product evaluation and purchase decision, in: International Textile and Apparel Association (ITAA) Annual Conference Proceedings, volume 75, 2018, p. 3. [21] S. T. Tong, E. F. Corriero, R. G. Matheny, J. T. Hancock, Online daters’ willingness to use recommender technology for mate selection decisions, in: IntRS@RecSys, 2018. [22] P. Reddy, K. Chaudhary, B. Sharma, R. Chand, The two perfect scorers for technology acceptance, Education and Information Technologies 26 (2021) 1505–1526. doi:10.1007/ s10639-020-10320-2. [23] A. Cypher, Eager: programming repetitive tasks by example, Proceedings of the SIGCHI
Conference on Human Factors in Computing Systems (1991). [24] T. Masui, K. Nakayama, Repeat and predict—two keys to eficient text editing, in: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, CHI ’94, Association for Computing Machinery, New York, NY, USA, 1994, p. 118–130. URL: https://doi.org/10.1145/191666.191722. doi:10.1145/191666.191722. [25] J.-D. Ruvini, C. Dony, Ape: Learning user’s habits to automate repetitive tasks, in: Proceedings of the 5th International Conference on Intelligent User Interfaces, IUI ’00, Association for Computing Machinery, New York, NY, USA, 2000, p. 229–232. URL: https: //doi.org/10.1145/325737.325854. doi:10.1145/325737.325854. [26] E. Horvitz, J. Breese, D. Heckerman, D. Hovel, K. Rommelse, The lumière project: Bayesian user modeling for inferring the goals and needs of software users, UAI’98, Morgan Kaufmann Publishers Inc., San Francisco, CA, USA, 1998, p. 256–265. [27] N. Yorke-Smith, S. Saadati, K. L. Myers, D. N. Morley, Like an intuitive and courteous butler: A proactive personal agent for task management, AAMAS ’09, International Foundation for Autonomous Agents and Multiagent Systems, Richland, SC, 2009, p. 337–344. [28] B. Peintner, J. Dinger, A. C. Rodriguez, K. L. Myers, Task assistant: Personalized task management for military environments, in: K. Z. Haigh, N. Rychtyckyj (Eds.), Proceedings of the Twenty-First Conference on Innovative Applications of Artificial Intelligence, July 14-16, 2009, Pasadena, California, USA, AAAI, 2009. URL: http://aaai.org/ocs/index.php/ IAAI/IAAI09/paper/view/236. [29] A. Bandura, Social foundations of thought and action: A social cognitive theory, Prentice
Hall, Englewood Clifs, NJ, 1986. [30] J. Cacioppo, R. Petty, The need for cognition, Journal of Personality and Social Psychology 42 (1982) 116–131. doi:10.1037/0022-3514.42.1.116. [31] J. B. Rotter, Generalized expectancies for internal versus external control of reinforcement.,
Psychological monographs 80 1 (1966) 1–28. [32] C. S. Dweck, C. yue Chiu, Y. yi Hong, Implicit theories and their role in judgments and reactions: A word from two perspectives, Psychological Inquiry 6 (1995) 267–285. URL: https://doi.org/10.1207/s15327965pli0604_1. doi:10.1207/s15327965pli0604\_1. arXiv:https://doi.org/10.1207/s15327965pli06041. [33] F. Carmagnola, F. Cena, C. Gena, User modeling in the social web, in: B. Apolloni, R. J. Howlett, L. C. Jain (Eds.), Knowledge-Based Intelligent Information and Engineering Systems, 11th International Conference, KES 2007, XVII Italian Workshop on Neural Networks, Vietri sul Mare, Italy, September 12-14, 2007, Proceedings, Part III, volume 4694 of Lecture Notes in Computer Science, Springer, 2007, pp. 745–752. URL: https://doi.org/10. 1007/978-3-540-74829-8_91. doi:10.1007/978-3-540-74829-8\_91. [34] S. Berkovsky, T. Kuflik, F. Ricci, Cross-representation mediation of user models, User Model.
User Adapt. Interact. 19 (2009) 35–63. URL: https://doi.org/10.1007/s11257-008-9055-z. doi:10.1007/s11257-008-9055-z. [35] F. Cena, C. Gena, C. Mattutino, M. Mioli, B. Treccani, F. Vernero, M. Zancanaro, The impact of psychological traits and rule recommendations in a smart home context: an empirical study (under review).