Reducing Algorithmic Bias in Generative Artificial Intelligence-Based Cyberbullying Detection Systems | IJCT Volume 13 – Issue 1 | IJCT-V13I1P29
International Journal of Computer Techniques
ISSN 2394-2231
Author
Ms Pooja Banerjee, Neeraj Kumar
Abstract
The explosive growth of social media has put more pressure on the issue of cyberbullying and its effect on the well-being of its users. Although definitive solutions have become common, using artificial intelligence-inspired detection systems to address harmful content to a moderate degree, there is growing evidence to suggest that they tend to be algorithmically biased, with a disproportionate rate of misclassification occurring when applied to linguistic variations that align with a certain demographic or cultural group. This defeats equity, confidence, and psychological security on the internet. This paper suggests a generative artificial intelligence framework to improve cyberbullying detection, in addition to the proactive reduction of bias. The model combines language modelling in contexts based on transformer-based generative representations and fairness aware optimization. Balanced data sampling, counterfactual data augmentation and loss functions that have fairness constraints are used as a bias reduction measure applied during model training. A dataset of multi-source cyberbullying composed of various linguistic phrases is experimentally tested. The measures of performance are accuracy, precision, recall, and F1 score, as well as having fairness metrics such as demographic parity difference and equal opportunity difference. Findings show that the given approach is competitive in terms of classification performance and its inter-group bias is lower than in the case of the baseline deep learning models. The results point to the significance of ethical and equity concerns in generating artificial intelligence systems of content moderation. The suggested framework will help to create inclusive, responsible, and safe psychological online spaces.
Keywords
Detection of cyberbullying, mitigation of algorithmic bias, fairness in machine learning, Natural language processing, transformer models, ethical artificial intelligence.
Conclusion
This paper presented an equal opportunity built artificial intelligence system of cyberbullying. In contrast to the traditional detection methods, which care mainly about predictive performance, the given approach will actively include the fairness requirements into the optimization task: The framework guarantees the reduction of bias is no longer a post hoc fix as the demographic parity and equal opportunity regularization will both be applied during training as it is a core goal of learning as well. Experimental analysis supported that generative transformer based contextual embeddings are useful in enhancing classification over baseline convolutional architectures. More to the point, fairness regularisation caused a critical decrease of the demographic disparity as the difference of Demographic Parity dropped to 0.05 and the competitive F1 scores were kept at the same level. The empirical trade-off curve verified that it is not the mutually exclusive objective but together fairness and predictive accuracy can be optimised by controlled regularisation. The results prove the possibility of ethical congruence and performance effectiveness of generative artificial intelligence systems in content moderation. The flexibility of the fairness coefficient is pronounced, and the possibility to adapt to the environment with different regulatory or ethical demands. Regardless of these contributions, there are still a number of limitations. The evaluation of fairness depends on the proper demographic annotation, which is not necessarily easily obtained. Moreover, the present research considers fairness as a major aspect of binary demographic partitions. Future studies ought to consider the subject of multi-group fairness constraints, multilingual datasets, and adaptive fairness weighting mechanisms. Transparency and user trust could also be enhanced by seeking further clarification on explainable fairness-aware models. In short, this framework can be used to further the creation of diverse, accountable, and contextually sound cyberbullying detection measures. Incorporating fairness as a principle of optimisation is a key challenge to developing credible generative artificial intelligence systems, which can act to enhance digital well-being (Mittelstadt, 2021) (al. E. F., 2020) (al. M. W., 2022).
References
1.K. Dinakar, R. R. (2011). Modeling the detection of textual cyberbullying. Proc. ICWSM.
2.V. Nahar, X. L. (2013). An effective approach for cyberbullying detection. Communications in Information Science and Management Engineering.
3.T. Brown et al. (2020). Language models are few-shot learners. NeurIPS.
4.al., A. R. (2021). Scaling language models: Methods and lessons learned,. ICML.
5.Guttag, I. S. (2021). A framework for understanding unintended consequences of machine learning,. arXiv.
6.al., R. B. (2021). On the dangers of stochastic parrots. FAccT.
7.al., M. A.-G. (2020). Deep learning for cyberbullying detection: A systematic review,. IEEE Access.
8.al., R. R. (2021). Transformer-based approaches for cyberbullying detection. Applied Sciences.
9.A. Mishra et al. (2021). BERT-based cyberbullying detection in social media. IEEE Access.
10.T. Wolf et al. (2020). ransformers: State-of-the-art NLP. EMNLP.
11.al., A. M. (2021). A survey on bias and fairness in machine learning. ACM Computing Surveys.
12.S. Dev et al. (2020). Mitigating gender bias in NLP models,. ACL.
13.Roth, H. C. (2020). A snapshot of the frontiers of fairness in ML,. Communications of the ACM.
14.Haas, B. C. (2020). Fairness in machine learning: A survey,. ACM Computing Surveys,.
15.Calders, F. K. (2020). Fairness aware machine learning in practice. Data Mining and Knowledge Discovery,.
16.al., S. G. (2021). Fairness in NLP: Systematic review,. ACL Findings.
17.al., A. T. (2020). Disparities in toxicity detection across dialects. ACL.
18.Gebru, R. B. (2020). Gender shades revisited,. ACL.
19.al., Y. R. (2021). Fairness-aware machine learning systems. IEEE software.
20.Mittelstadt, B. (2021). Principles alone cannot guarantee ethical AI,. Nature Machine Intelligence.
21.al., E. F. (2020). AI ethics guidelines global review. Nature Machine Intelligence.
22.al., M. W. (2022). Ethical and societal implications of generative . IEEE .
How to Cite This Paper
Ms Pooja Banerjee, Neeraj Kumar (2025). Reducing Algorithmic Bias in Generative Artificial Intelligence-Based Cyberbullying Detection Systems. International Journal of Computer Techniques, 12(6). ISSN: 2394-2231.
Reducing Algorithmic Bias in Generative Artificial Intelligence-Based Cyberbullying Detection SystemsDownload
Related Posts:
