This dissertation explores the orchestration and evolution of Virtual Learning Companion (VLC), support as an add-on in both simulated and real social media environments. Furthermore, it investigates the impact of the VLC on users’ judgments, critical thinking, levels of agreement, and awareness of both non-toxic and toxic content, using a blend of qualitative and quantitative statistical methodologies. Before shaping the VLC system and its potential scenarios, we created a minigame that enables users to categorize internet images based on various dimensions such as hate speech, cyber mobbing, verbal violence, and discrimination. This game not only promotes critical thinking skills but also serves as a tool for learning through conflict. The data collected from the game provide invaluable insights into students’ perceptions of harmful content and contribute to fostering responsible online behavior. Moreover, the application enhances users’ ability to analyze and evaluate online content. To address any potential controversies, we examined various methodologies for assessing agreement levels and selected the most appropriate one for calculating this index among raters. This chosen methodology was also applied to gauge the level of agreement among learners in the context of image manipulation within the VLC scenario. Learning Companion Systems (LCS) enhance traditional Intelligent Tutoring Systems (ITS) by incorporating a computer companion that offers personalized guidance and interactivity. This results in an enriched learning experience. Referred to as 'learning companions,' these intelligent and autonomous robots or virtual conversational agents are capable of maintaining long-term user relationships. They find applications across various domains, including education and business, and focus on specific tasks such as teaching math and programming or assisting users in online transactions. Our VLC system is designed as a Chrome plugin and as a simulated version of Instagram with modular functionalities. This design allows for flexible application across different scenarios. We also crafted a series of statements in xAPI standard format, specifically designed to log user interactions. We conducted empirical studies in Germany and Italy, focusing on daily social network issues such as hate speech, discrimination, racism, conspiracy theories, fake news, and especially image manipulation. The VLC browser plugin employs various chat dialogues to facilitate interactive learning experiences. It actively engages learners by posing knowledge-activating questions and providing pre-defined, insightful responses in certain cases. In the initial scenario, a qualitative study was carried out with a limited user group. Feedback from users led to the optimization of both the VLC system and the simulated environment, influencing the design and user interaction for better qualitative outcomes. Subsequently, a second study was conducted with a larger sample, providing quantitative data that corroborated our initial findings. Here, we honed in on image manipulation that could be addressed by reverse image search engines, offering significant insights. The results of the analysis from the German school indicate that the time learners spent engaging with VLC recommendations from reverse image search and metadata correlated with improved judgment. Comparative studies were also conducted in Italy to discern the impact of cultural variables, reinforcing the VLC’s effectiveness in different educational settings for image manipulation scenario. These investigations confirmed the VLC’s consistent influence on fostering critical evaluation skills, irrespective of cultural background. Moreover, we also executed a targeted study dealing with racism through the VLC system in a simulated social media platform. This study unveiled the VLC’s potential and adaptability to educating young minds about the adverse effects of racism, thereby contributing to a more inclusive society.