Giuseppe Fenza

University of Salerno · Italy

Unclaimed · AI-structured profileORCID OpenAlex

Computational social scientist applying LLM-based multi-agent simulation to opinion dynamics and digital twins.

papers here

reviewed by Evolve

first-author

corresponding

—

works · OpenAlex

h-index · OpenAlex

Bibliometric figures come from OpenAlex (2026-10-15) and cover the author's entire public record, not just the papers listed here. Reproducibility signal comes from Evolve's own reviews.

Research profile

Output—

Not yet scored: this snapshot lacks a verified full publication list, so the output axis — total works, annual productivity, and venue spread — cannot be computed. Partial signals (33 works in the last three years) await reconciliation.

Impact—

Not yet scored: the citation aggregates required for the impact axis are not available in this snapshot. It does report an h-index of 25 and i10-index of 51, but the composite impact score is uncomputed and is not estimated.

Focus0/100

Scores 0/100 on focus. The recent field split is actually tight — LLM-based multi-agent simulation, opinion dynamics with social network analysis, and LLM-based social simulation each take about 33% — but topic counts show breadth across a long career: Semantic Web and ontologies (36 works), misinformation (18), rough sets and fuzzy logic (14), hate speech detection (13), recommender systems (11), which drives the dispersion.

Collaboration39/100

Scores 39/100. The snapshot counts 6 distinct coauthors, a comparatively small direct network. Recurring names are Omran Berjawi and Rida Khatoun, with Sherali Zeadally also frequent. In the three cataloged 2026 preprints Fenza appears in middle position twice and last position once, none as corresponding author, consistent with senior co-authorship on team-driven simulation projects.

Activity100/100

Scores 100/100, the profile's strongest axis. The record shows 33 works in the last three years and a rising curve: 12 works in 2024, 13 in 2025, and 8 already indexed in 2026. Annual counts also peaked at 15 in 2021 and 13 in 2020, so the high cadence is sustained rather than new.

Reproducibility55/100

Scores 55/100, a mean over 3 Evolve-reviewed papers, each rated medium reproducibility readiness. Verdicts flag concrete gaps: the digital-twin study ships no code and uses sentiment-proxy ground truth on small networks; the bias-amplification study is single-model and one-sided; the influencer study leaves significance untested with thin baselines.

Research summary

Giuseppe Fenza is a researcher at the University of Salerno, Italy, whose current program centers on computational social science, specifically LLM-based social simulation and opinion dynamics. Three 2026 arXiv preprints in Evolve's catalog illustrate the pivot: a generative LLM digital-twin framework for opinion dynamics on social networks; a study of how a biased LLM minority (at most 20% of agents) shifts neutral opinions beyond a Friedkin-Johnsen baseline; and an influencer-ranking study that uses the FJ fundamental matrix to simulate opinion shifts on 2020 US-election tweets, on which he is last author. The bibliometric snapshot, resolved from OpenAlex, shows long-standing breadth — Semantic Web and ontologies (36 topic-assigned works), misinformation (18), rough sets and fuzzy logic (14), hate speech detection (13), and recommender systems (11) — now converging on agent-based social simulation. Activity is the standout dimension (100/100): 33 works in the last three years, with 12 in 2024, 13 in 2025, and 8 already indexed in 2026. The snapshot reports an h-index of 25 and an i10-index of 51, though output and impact dimension scores are not yet computed. Reproducibility is moderate (55/100) across the three reviewed papers, each rated medium readiness, with verdicts citing missing code, proxy ground truth, and thin baselines. Collaboration is comparatively narrow (39/100; 6 distinct coauthors), anchored by Omran Berjawi, Rida Khatoun, and Sherali Zeadally.

Collaboration pattern

The snapshot records 6 distinct coauthors, consistent with the moderate collaboration score of 39. Recurring partners in the record are Omran Berjawi and Rida Khatoun, with Sherali Zeadally also frequent. Across the three cataloged 2026 preprints Fenza holds middle authorship twice and last authorship once, with no corresponding role, suggesting senior co-authorship on team-led simulation projects rather than solo or corresponding leadership.

Research keywords

opinion dynamicsLLM-based social simulationmulti-agent systemsgenerative agentsdigital twinFriedkin-Johnsen modelsocial network analysisinfluencer identificationbias amplificationmisinformationsentiment analysisSemantic Web and ontologies

Highlights

LLM digital twins for opinion dynamics claim >50% MAE gains
A generative framework builds LLM-based digital twins of social network users; Evolve notes the claimed >50% MAE improvement over classical opinion models on two Twitter datasets, while flagging small networks, sentiment-proxy ground truth, and no released code.
A ≤20% biased LLM minority bends whole-network opinion
Multi-agent simulations show a biased LLM minority of at most 20% shifts neutral agents beyond a Friedkin-Johnsen baseline, with generated rhetoric decoupled from population shares; Evolve calls the result plausible but notes the single-model, one-sided design.
Ranking influencers with the Friedkin-Johnsen fundamental matrix
Last-authored by Fenza, the study ranks influencers via the FJ fundamental matrix and simulates opinion shifts on 2020 US-election tweets; Evolve finds the direction promising but notes untested significance and thin baselines.

Timeline

2006Earliest indexed activity in the record: 1 work, starting a publication history now spanning two decades.
2011Early peak of the first decade: 9 works, the highest annual count of the 2006–2013 phase (7 in 2009).
2020Output surge begins: 13 works in 2020, followed by the record maximum of 15 in 2021.
2025Sustained high cadence: 12 works in 2024 and 13 in 2025, approaching the historical maximum.
2026Pivot to LLM-based social simulation: 8 works indexed so far, including three arXiv preprints in Evolve's catalog on digital twins, bias amplification, and influencer persuasion.

Papers in the Evolve catalog

Frequent co-authors

Omran Berjawi×3 Rida Khatoun×3 Omran Berjawi×1 Giuseppe Fenza×1 Rida Khatoun×1 Sherali Zeadally×1

About this profile

This profile is AI-structured by Evolve.Pub from public bibliographic data (OpenAlex) and Evolve's own paper reviews. It is not written or endorsed by the researcher unless the profile is claimed. Radar dimensions are computed from real bibliometrics with fixed formulas — raw figures are shown alongside; missing dimensions mean insufficient data, never zero. Errors? Claim the profile to correct or hide it.

Identity is resolved from OpenAlex bibliographic data (ORCID 0000-0002-4736-0113); profile scope is limited to the bibliometric snapshot plus Evolve's reviews of 3 preprints.
h-index (25) and i10-index (51) are snapshot-level counts; the output and impact dimension scores remain uncomputed.
Reproducibility mean (55) rests on only 3 reviewed arXiv preprints; annual counts have gaps (2014 and 2018 absent) and 2026 is a partial year.
Collaboration metrics (39; 6 distinct coauthors) come from a narrow snapshot window and may understate the full co-author network.