Postdoctoral Researcher

The Ubiquitous Internet Unit of IIT-CNR (Pisa, Italy) is scouting for talented researchers at the post-doctoral level. We are collecting expressions of interest for the following research topic:

• Signed graphs for online social networks: data-driven characterization, temporal dynamics, applications

The position is open at postdoctoral level, with the research direction adaptable to the expertise and interests of the applicant. Candidates with strong analytical and computational skills, as well as an interest in social network analysis, AI/ML, and data-driven research, are encouraged to apply.

Candidate profile:

• PhD in Computer Science, Physics (complex networks), or Mathematics.
• Strong programming skills for data collection, processing, and model training.
• Experience or interest in interdisciplinary research, such as computational social science.

On the research topic

Since Granovetter’s seminal work on the strength of social ties, we have understood that social relationships are shaped by multiple factors, including frequency of interaction, trust, and emotional connection. Most research on online social networks has focused on interaction frequency, as it is the easiest metric to compute and quantify. However, recent studies have shifted toward more qualitative aspects of relationships, moving beyond mere interaction counts to understanding the nature of social connections.

One of the most widely used ways to encode the quality of relationships is by assigning a positive or negative sign to them. Positive relationships often represent trust, homophily, or mutual support, while negative relationships may indicate conflict, distrust, or rivalry.

Everyday experience shows that information does not flow equally through positive and negative ties—for instance, people may avoid sharing personal details with those they do not trust. This fundamentally challenges traditional models of social communities, information diffusion, and network structure. When social networks are represented as graphs, these signs are attached to the links between nodes. While emerging properties like information diffusion, community formation, and polarization have been largely investigated by modeling social networks as unsigned graphs, the effect of the polarity (positive or negative) of ties on these processes is still largely unexplored.

Despite the importance of signed relationships, a major limitation in the field is the lack of reliable ground truth datasets. The few existing signed datasets are often outdated and domain-specific (e.g., product ratings, online reviews), making them unsuitable for studying general-purpose notions of positive and negative relationships in diverse social settings.

This research position aims to fill this gap by designing and deploying new methodologies to

• collect ground truth signed graphs
• use it to train AI/ML models and improve social network analysis.

Possible approaches include:

• Conducting surveys and experiments to gather data, either through direct administration or via crowdsourcing platforms.
• Leveraging large language models (LLMs) such as ChatGPT to infer relationship signs at scale.
• Developing novel, non-ML-based methods for inferring signed relationships from collected social interactions data.
• Applying the collected datasets and derived models to improve downstream tasks, such as community detection, information diffusion modeling, and polarization analysis.

Funding and partnerships

The activities of this topic will be supported by:

• FAIR: Extended Partnership on Artificial Intelligence (funded by the National Recovery and Resilience Plan (NRRP), European Union - NextGenerationEU)
• ICSC: National Research Centre for High-Performance Computing, Big Data and Quantum Computing