|Talks|

Decoding the Hidden Knowledge: Uncovering Latent Attributes in Entities Using Information Theory

Hybrid
Past Talk
Ossi Mokryn
Senior Lecturer, University of Haifa
Mar 1, 2023
2:00 pm
Mar 1, 2023
2:00 pm
In-person
4 Thomas More St
London E1W 1YW, UK
The Roux Institute
Room
100 Fore Street
Portland, ME 04101
Network Science Institute
2nd floor
Network Science Institute
11th floor
177 Huntington Ave
Boston, MA 02115
Network Science Institute
2nd floor
Room
58 St Katharine's Way
London E1W 1LP, UK

Talk recording

"He who knows, and knows not that he knows, is asleep; wake him," a quote from the Persian philosopher Ibn Yami, highlights the importance of "unknown knowns" -- knowledge that exists but is not acknowledged -- in decision-making, risk assessment, and fields such as psychology and cognition, as well as its role in the explanation of biases. Building on the observation of Sigmund Freud that "the missing, not conveyed, is an integral part of the whole," we propose a methodology for identifying latent, unknown-known attributes of entities. Our approach, called Population-based Latent Personal Analysis (LPA), uses an information-theoretic approach to find population-based attributes for entities. By identifying elements whose Shannon information differs most from their counterparts in the population, LPA can determine an entity's locally overused rare population elements, and elements that are locally underused or missing that are popular in the population. These elements contribute to the entity's LPA's distance attribute and determine its LPA's signature attribute. We show that LPA's attributes enable a quantitative and qualitative understanding, and that the unknown knowns provide additional important information. We demonstrate the practical applications of LPA attributes in various areas, such as impersonation detection in social media, the spectral spread of sub-repertoires of clonal B-cell populations within a person, and the digital humanities. We also discuss how these attributes can be used for identifying trends and patterns in temporal data.

About the speaker
Osnat (Ossi) Mokryn is a tenured senior lecturer in the Department of Information Systems at the University of Haifa, Israel, and a visiting scholar at the Harvard John A. Paulson School of Engineering and Applied Sciences. She heads the Content & Social Networks group (SCANLab) that focuses on understanding complex interaction systems using real-world data, with a specific focus on human-AI interactions. Ossi develops methods based on information theory and cognition for extracting knowledge from unstructured and temporal data, and her recent work on identifying emotional experience from text has received an award from the Israeli Association for Information Systems. During the pandemic, her research focused on exploring the interplay between the dynamics of a propagating viral process and the underlying real-world interaction network. Currently, she researches determinants of expertise in research and perceptions of rankings and hierarchy.
Share this page:
Mar 01, 2023