Exploring the Power of Dual Deep Learning for Fake News Detection
Bibliographic record
Abstract
The rise of social media has intensified the spread of fake news, a problem further exacerbated by generative artificial intelligence (AI). Hence, the need for improved detection of both human-created and AI-generated fake news using advanced AI models is critical. This paper proposes a survey to assess knowledge and attitudes towards news and AI, combining demographic data, personality traits, and the ability to distinguish between real and AI-generated news. Additionally, we create a new dataset, ERAF-News, containing real, fake, AI-generated true, and AI-generated fake news. To classify different types of news, we developed a dual-stream transformer model, DuSTraMo. This model leverages the capabilities of two parallel transformers to enhance the accuracy of news classification. The survey, involving 83 participants from 9 countries, revealed that respondents struggle to differentiate human-generated from AI-generated news. Notably, BERT outperformed GPT-2 and BART in generating realistic text, and RoBERTa and DistilBERT achieved over 98% accuracy in fake news classification. Dual-GPT models also showed high accuracy.This study underscores the effectiveness of the DuSTraMo model and the ERAF-News dataset in enhancing the detection of both human-created and AI-generated fake news. The findings highlight the increasing dominance of AI in this domain and the pressing need for advanced methods to combat fake news. Additionally, a survey examining users’ responses to fake news reveals a concerning inability to accuratelyidentify false information.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.008 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.001 | 0.002 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".