Our micro-TWONy demonstrator shows the impact of recommendation algorithms of online social networks on the emotionality of users‘ feeds. The prototype allows users to choose, whether posts in a feed should simply be ranked chronologically, or whether they should be ranked by an algorithm, that rewards emotional posts. After making the choice for a recommendation algorithm, TWONy-users can simulate a feed with the help of autonomous generative agents. They can then observe, which effect the chosen recommendation algorithm has on the emotionality of the resulting feed: the visualized network metrics show, whether positive or negative emotions prevail in the feed and how emotionality evolves over time.
More technically speaking: TWONy is a micro-simulation prototype designed to show the impact of online social network (OSN) mechanics — particularly recommendation algorithms — on emotional contagion and discourse dynamics. By harnessing the capabilities of Large Language Models (LLMs), the system generates an ecosystem of synthetic, politically engaged digital personas that interact within a controlled social media environment. These autonomous agents emulate human behaviors through in-context prompting techniques, enabling users to observe emotional transmission patterns under systematically varied conditions while circumventing the constraints inherent to real-user experimentation. The prototype implements two distinct recommendation paradigms: a baseline chronological feed and an emotion-prioritizing ranking mechanism that amplifies content based on emotional intensity, allowing the examination of the formation and reinforcement of echo chambers. Emotional valence is quantified via a fine-tuned BERT model, while network-level and agent-level visualizations track emotional cascades and polarization dynamics throughout the lifecycle. The system is architected as a browser-based application leveraging modern web technologies and decentralized APIs.
How-to micro TWONy:
By default, you see the „News Feed“ page first. As a user, you can type in a post, post it to the feed (by clicking on the blue „Post“ button). Subsequently, you can start a simulation based on your own post (by clicking on the green „Start“ button). You can also choose the Language Model, which is to be used for the simulation (Llama, Mistral, Deepseek). The resulting feed is populated by generative agents, which emulate human behavior. Under each post, the positive (joy, optimism, trust) and negative (anger, fear, pessimism) emotions are displayed.
On the right of the page, under „Network Metrics“, a measure of the overall positive and negative emotions in the feed is displayed over time. Under „User Metrics“, the positive and negative valence per user is displayed. You, as the author of the initial post are the “Human”. The others, such as “ProgressiveRage” and “FactChecker”, are the generative agents. To the left of the page, you can access the menu (News Feed, Ranking Settings, Agent Settings). Under “Ranking Settings”, you can choose the recommendation algorithm, with which you want to run the simulation. You can choose between a plain chronological ranker and an emotion-based ranking. For the emotion-based ranking you can choose, whether negative and positive emotionality of posts, should be rewarded (weighting increased) or not (decreased weighting). These settings are then displayed to the bottom right of the „News Feed“ page, under the „User Metrics“.
Under “Agent Settings”, you can see, how the generative agents in the feed are prompted, and customize the instructions, if you wish to. Under „Customize Personas“, you can view the prompted humor style, communication pattern, emotional expression, values and beliefs, interests and hobbies, social interactions, personality traits and cultural background for each generative agent. You can also customize them, add new personas and delete existing ones for your simulation.