cs.AIFeb 28, 2026

Agent-Based User-Adaptive Filtering for Categorized Harassing Communication

Authors: Zenefa Rahaman, Sandip Sen

Organizations: The University of Tulsa, Tulsa, Oklahoma

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

We propose an agent-based framework for personalized filtering of categorized harassing communication in online social networks. Unlike global moderation systems that apply uniform filtering rules, our approach models user-specific tolerance levels and preferences through adaptive filtering agents. These agents learn from user feedback and dynamically adjust filtering thresholds across multiple harassment categories, including offensive, abusive, and hateful content. We implement and evaluate the framework using supervised classification techniques and simulated user interaction data. Experimental results demonstrate that adaptive agents improve filtering precision and user satisfaction compared to static models. The proposed system illustrates how agent-based personalization can enhance content moderation while preserving user autonomy in digital social environments.

Figures & tables

Figure 1. Individualized Agent between Social media platform and user helping to customize the filtering of communication to remove or flag online communication.
#Category# of tweets
0General harassment79
1Cruel statement1054
2Religious/racial/ethnic89
3Sexual orientation11
4Sex/ gender656
5Threat236
Table 1. Frequency of Tweet Categories
Figure 2. Percentage of data selected by users to be filtered for the different harassment categories.
Figure 3. Percentage of tweets selected to be filtered for each category and for perceived harassment intensity levels.
CategoryCategoryP-DiffLowerUpper
010.0704-0.21390.0759
020.0055-0.14480.1339
030.133-0.27910.0182
040.1370.014330.2546
050.0898-0.23420.0582
060.1142-0.25970.0357
Table 2. Confidence Interval Coefficient for % tweets filtered for Intensity Level 2 (columns highlighted denote statistically significant difference; 19 of 28 pairs are different).
Figure 4. Histogram of # users wanting to filter a certain number of tweets. X axis represents # of tweets (0-75), Y axis represents # of users wanting to filter that many tweets.
Figure 5. (a) Frequency plot for filtering choices (X-axis: # users, out of 5, who chose to filter a tweet, Y-axis: # of tweets), (b) User Choice variability plot.
Figure 6. An agent-based user-adapted harassment filter.
Figure 7. Accuracy of General Vs user-adapted Filters.
Figure 8. Majority class Vs NB, SVM, RF classifiers.
Majority VsSVMRFNB
Win14913252
Lose166143136
tie4283169
Table 3. Comparison of performance of the learning algorithm with the majority Algorithm.
More in this topic

Safety Filtering

Safety filtering intercepts actions or generated content to prevent harm while preserving as much useful behavior as possible.

See all 126 papers

Explore similar work

CardsList

  1. Who Decides What Is Harmful? Content Moderation Policy Through A Multi-Agent Personalised Inference Framework

    May 2, 2026Ewelina Gajewska, Michal Wawer, Katarzyna Budzynska +1LLM GuardrailsContent Moderation

    Warsaw University of Technology, Warsaw, Poland

  2. Moderation by Interrogation: Uncovering Concealed Intent in Agent Social Networks

    May 13, 2026Ali Al-Lawati, Nafis Tripto, Abolfazl Ansari +4Markov Chain Monte CarloIntent Classification

    The Pennsylvania State University, USA

  3. Simulating Hate Speech Cascades with Multi-LLM Agents: Empirical Grounding, Modeling Fidelity, and Intervention Strategies

    May 21, 2026Fan HuangAgent-Based ModelingMulti-Agent LLMs

    Indiana University Bloomington