Fuzzy Logic: A Fuzzy-based Privacy Recommender System for Political Participation


In the digitalised world, citizens – while having control over their personal information – are increasingly exposing their data, and there are plenty of online platforms ready to collect this.

Aigul Kaskina, in collaboration with the FMsquare Foundation, presents the concept of a fuzzy-based recommender system for user account privacy settings that can be used for citizen participation on online political platforms.

Read the original research: doi.org/10.1007/978-3-031-06021-2_1

Read more about the FMsquare Foundation in Research Outreach

Visit the FMsquare Foundation’s website: fmsquare.org


Image Source: Adobe Stock Images / mh.desing




Hello and welcome to ResearchPod. Thank you for listening and joining us today.

In this episode, we look at the work of Aigul Kaskina, who acquired the degree of Doctor of Philosophy in Computer Science from University of Fribourg, Switzerland for her research using the framework of ‘fuzzy logic’ – in collaboration with the FMsquare Foundation. Her work presents the concept of a fuzzy-based recommender system for user account privacy settings that can be used for citizen participation on online political platforms. Currently, Aigul is a Data & AI scientist at Accenture based in Zurich, Switzerland.

In the digitalised world, every piece of online personal information might become sensitive information. On the one hand, there are plenty of online platforms ready to collect citizens’ personal data. On the other, however, citizens – though having control over their personal information – are increasingly exposing their data.

People also seem to lack concern with privacy, until receiving an appropriate notice. Considering the complicated human’s privacy decision-making nature, the motivation of Kaskina’s thesis focuses on the analysis of people’s privacy behavior in the context of online political platforms, posing a research question: how do we quantify privacy attitudes and how do we generate privacy settings recommendations which might positively impact a person’s future desire for disclosure.

Aigul Kaskina proposes an information system design that supports citizens to regulate their privacy boundaries in the online environment for political participation. Her research work aims to develop a conceptual framework and intelligent engine that enhances citizens’ privacy in e-Democracy application by generating personalised privacy recommendations.

What’s more, the application of fuzzy logic techniques is proposed in order to facilitate the accuracy and utility of intelligent privacy recommendations. As a result of the research, two artifacts were developed: the first artifact – a citizen privacy profile framework and the second artifact – a fuzzy-based privacy recommender system prototype integrated into web-based application for political participation.

People may often feel unsure when deciding what to share online about themselves, often weighing up the potential benefits and risks of sharing information. Even if people state that they want privacy, their actions may suggest otherwise. Such inconsistent behavior is termed as a ‘privacy paradox’, which refers to the phenomenon where individuals express concerns about their privacy but still engage in behaviors that compromise it. It highlights the contradiction between people’s stated privacy preferences and their actual online activities that involve sharing personal information. As an example, people can easily give up their personal data, for instance, for the alimentary or monetary rewards by neglecting associated privacy risks and potential consequences.

Technology services could exploit users’ low privacy awareness by extensively collecting data, such as cell-phone location, click-streams, credit card usage, and more – by subsequently selling personal data to entities interested in obtaining the information. While physical privacy can be understood as a result of physical or tactical risk-avoidance behaviour, the online privacy instead is harder to understand, and therefore there is a greater uncertainty in decision-making. Given the complex nature of human decision-making regarding online privacy, this can encourage individuals to use cognitive shortcuts like heuristics helping to make a simplified decision-making strategy relying on cues such as popularity, rule of thumb or word of mouth.

Aigul’s research focused on a problem of citizens’ privacy behavior on the platform for political participation called Participa Inteligente. The Participa Inteligente platform was an academic project at the University of Fribourg that arose from the concern for citizen misinformation on policy statements –  allowing citizens to generate spaces for discussion and participation on topics of interest to society.

Moreover, the platform aimed to enhance civic participation and empowerment by providing different types of recommendations, such as: political topics, groups, articles, and users, among others. The platform was thereby expected to facilitate citizens’ discussions and debate. However, the extent to which citizens’ willingness to actively participate on the platform by disclosing political interests or opinions depended on the personal desire for disclosure.

An integral aspect of this research involves the development of a citizen privacy profile framework that introduces the concept of fuzzy privacy profiles. Privacy preferences might vary considerably between different people. To group people based on their privacy preferences the clustering techniques can be used.  Clustering is an unsupervised learning, where a clustering algorithm organizes a given set of objects into similar groups (clusters). For the citizen privacy framework Aigul Kaskina uses fuzzy-based clustering algorithms on the dataset of users’ privacy preferences.

From the study, she argues that nuances of user privacy preferences can be well detected with the fuzzy clustering based on the fuzzy set theory introduced by Lotfi Zadeh. It allows to represent a privacy preference of the user in a multidimensional form and assign it to several groups to some certain degree of belonging. This approach can reduce an oversimplification problem on estimating users’ privacy preferences and improve the estimation accuracy. To the best of Kaskina’s knowledge, this was a first attempt to interpret the multidimensionality of user privacy preferences using a fuzzy logic approach.

This research offered a functional space which allowed users to express their desired level of privacy by explicitly defining privacy settings for their accounts. Specifically, the designed citizen privacy profile framework determined essential components for building a privacy management tool for an online platform, where the framework’s components were customised and implemented to meet the needs of the Participa Inteligente platform. Also, this research proposed the conceptual design and implementation of a fuzzy-based privacy settings recommender system, which applied model-based approaches using a real-world dataset of users’ privacy settings. The recommender system calculates privacy recommendations using the derived information from fuzzy privacy profiles mentioned earlier.

The fuzzy-based recommender system concentrates beyond the multidimensional structure of privacy preferences thanks to the used approach of fuzzy-based clustering techniques. The privacy recommendations were afterwards evaluated by the users to measure their perceived usefulness or satisfaction with a received privacy recommendation. The integration of a fuzzy-based recommender system aimed to preserve citizens’ privacy on an automatic basis and facilitated their participation on political matters by releasing their cognition load from the burden of privacy decision-making.

The utmost utility of both artifacts was focused on users of the platform, such that by analysing their privacy behavior to provide an optimised solution to their privacy settings configurations. Thus, the central part of the provided solution are the citizens. However, the optimisation goals might differ depending on the context of the platform. As an example, the system may not fully automate users’ privacy decision-making, but instead, determine a reasonable balance between automation and user’s manual control.

Also, it is difficult and somewhat unfair trying to associate individuals with a particular cluster or group – due to the inherent vagueness and uncertainty in their attitudes and preferences. Applying fuzzy logic facilitates in dealing with humans’ perceptions and cognitive estimations, as well as human imprecision, thereby providing more accurate and personalised solutions.

In some situations, privacy automation can go against the user empowerment. Automation can hold certain risks in terms of its validity and/or ensuing consequences in case of a recommendation adoption. If the user’s actual privacy behavior deviates from their privacy attitudes, then a system model based on this fallacious data can be considered unreliable. In this situation, it becomes a disadvantage to use static privacy framework to model user privacy modes using a fuzzy-based approach. Static models have a higher risk to overlook the “attitude-behavior” gap issue caused by the privacy paradox. To improve privacy models, dynamic cues, such as psychological or sociological aspects of users can be considered.

: ‘the application of fuzzy-based techniques to analyse user privacy preferences and the introduction of fuzzy privacy profiles is the first attempt done by this work. Moreover, no previous studies focused on the user-centric evaluation of privacy recommendations. However, this invites future research to tackle the users’ perceptions and adoption towards privacy recommendations and to focus on dynamic techniques for calculating privacy recommendations.’

That’s all for this episode, thanks for listening. Links to Kaskina’s research, including the book ‘Citizen Privacy Framework Case of a Fuzzy-based Recommender System for Political Participation’, based on her doctoral dissertation, can be found in the shownotes for this episode. And, as always, stay subscribed to ResearchPod for more of the latest science!

See you again soon.

Leave a Reply

Your email address will not be published.

Researchpod Let's Talk

Share This

Copy Link to Clipboard