Balancing Societal Impact and Digital Footprint: A Sustainability-Aware Framework for AI-Based Misinformation Detection
Journal of Sustainable Software Engineering and Information Systems · 19.11.2026 · 2 (2) · 167-183
DOI: 10.58723/jsseis.v2i2.261 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Sustainable Software Engineering and Information Systems, online erschienen am 19.11.2026. Verfasst von Sofowora Mayowa Abiola und 4 weiteren Autor:innen.
Abstract (Originalsprache)
Background
AI-based misinformation detection can protect digital information environments, but existing evaluations often separate predictive performance from energy consumption, inference efficiency, operational cost, and societal consequences. This limits assessments of deployment sustainability.
Aims
This study develops a sustainability-aware framework integrating societal benefits and digital footprint to identify deployment configurations balancing misinformation-exposure reduction, information-trust support, predictive performance, and computational efficiency.
Methods
A controlled experiment used 983 unique public news titles from the PolitiFact subset of FakeNewsNet. A word- and character-level TF-IDF classifier was evaluated under four configurations: individual full-precision inference, batched full-precision inference, batched INT8 inference, and compact INT8 edge/offline inference. Metrics included macro-F1, misinformation-exposure reduction, an operational information-trust support proxy, latency, throughput, estimated energy consumption, cloud cost, and serialized model size. Confidence intervals were estimated using 3,000 bootstrap resamples.
Results
Individual full-precision inference achieved a macro-F1 of .873, 88.8% estimated exposure reduction, and an 87.4% social-benefit score, consuming 3.14 Wh per million inferences. Batching preserved predictive and social outcomes while reducing energy consumption to 0.33 Wh. Batched INT8 maintained these outcomes, consumed 0.32 Wh, and reduced model size from 128.2 to 82.2 KB. Compact edge inference consumed 0.05 Wh and reduced model size to 14.5 KB, but macro-F1 declined to .847 and exposure-reduction potential to 83.2%.
Conclusion
Batched INT8 inference offered the strongest overall sustainability balance, preserving societal benefits while reducing energy consumption, latency, cloud cost, and model size. The framework supports integrated social–technical evaluation of sustainable misinformation-detection systems.
Zitieren
APA
Abiola, S. M., Norda, N. P., Akolade, M. T., Nakka, R., & Chandolu, S. B. (2026). Balancing Societal Impact and Digital Footprint: A Sustainability-Aware Framework for AI-Based Misinformation Detection. Journal of Sustainable Software Engineering and Information Systems, 2(2), 167-183. https://doi.org/10.58723/jsseis.v2i2.261
Vancouver
Abiola SM, Norda NP, Akolade MT, Nakka R, Chandolu SB. Balancing Societal Impact and Digital Footprint: A Sustainability-Aware Framework for AI-Based Misinformation Detection. Journal of Sustainable Software Engineering and Information Systems. 2026;2(2):167-183. doi:10.58723/jsseis.v2i2.261
BibTeX
@article{Abiola2026,
author = {Abiola, Sofowora Mayowa and Norda, Niken Parnia and Akolade, Mojeed Taiwo and Nakka, Rajeswari and Chandolu, Suresh Babu},
title = {Balancing Societal Impact and Digital Footprint: A Sustainability-Aware Framework for AI-Based Misinformation Detection},
journal = {Journal of Sustainable Software Engineering and Information Systems},
year = {2026},
volume = {2},
number = {2},
pages = {167--183},
doi = {10.58723/jsseis.v2i2.261}
}
EUVIMED