Results of the summarizing process for the arXiv paper: 2101.08035v1

Ontologies play a crucial role in influencing data analytics by determining what information is fed into algorithms. While algorithmic bias is a well-known concept, it is important to understand how bias manifests itself in ontologies that serve as a structuring mechanism for algorithm inputs. This study aims to examine and identify different types of bias that are relevant to ontologies, whether they are explicit or implicit. The researchers illustrate these eight types of bias using examples from existing production-level ontologies and literature samples. By doing so, they shed light on the sources of bias within ontologies. Furthermore, the researchers assess three COVID-19 ontologies that were concurrently developed to analyze the presence of bias. They find that each ontology exhibits different subsets of bias types to varying degrees. This assessment contributes to raising awareness about ethical considerations related to ontologies, particularly regarding the representation of information and knowledge. Overall, this preliminary assessment provides valuable insights into the nature of bias in ontologies and highlights the need for further research and attention towards ethical aspects in ontology development.

• - Ontologies play a crucial role in influencing data analytics
• - Bias can manifest itself in ontologies that serve as a structuring mechanism for algorithm inputs
• - The study aims to examine and identify different types of bias in ontologies, whether explicit or implicit
• - Eight types of bias are illustrated using examples from existing ontologies and literature samples
• - Three COVID-19 ontologies were assessed for the presence of bias, with each exhibiting different subsets to varying degrees
• - The assessment raises awareness about ethical considerations related to ontologies and representation of information and knowledge
• - The preliminary assessment provides valuable insights into the nature of bias in ontologies
• - Further research and attention towards ethical aspects in ontology development is needed.

Ontologies are important for analyzing data. Bias can show up in ontologies, which are like a structure for algorithms. The study wants to find different types of bias in ontologies, whether obvious or hidden. They looked at examples from existing ontologies and literature to find eight types of bias. They also checked three COVID-19 ontologies for bias and found different amounts in each one. This shows that we need to think about ethics when making ontologies. The study gives us good ideas about how bias can be in ontologies, but more research is needed." Definitions- Ontology: A way to organize information and knowledge. - Bias: When something is unfair or not balanced. - Algorithms: A set of rules that computers follow to solve problems. - Explicit: Something that is clear and easy to see. - Implicit: Something that is not directly stated but still there. - Ethics: Knowing what is right and wrong, and doing the right thing. - Development: Creating or making something better over time.

Ontologies and Algorithmic Bias: Examining the Presence of Bias in Ontologies

The use of algorithms to process data has become increasingly popular in recent years, with applications ranging from facial recognition technology to automated decision-making systems. While algorithmic bias is a well-known concept, it is important to understand how bias manifests itself in ontologies that serve as a structuring mechanism for algorithm inputs. This article examines the presence of bias in ontologies and provides insights into the nature of bias within them.

What are Ontologies?

An ontology is an organized set of concepts and relationships used to describe a domain or knowledge area. In other words, it is a way of representing information about the world around us in a structured format. Ontologies can be used to represent facts about people, places, objects, events, etc., which can then be used by algorithms as input data for analysis and decision-making processes.

Types of Bias Found Within Ontologies

A study conducted by researchers at Stanford University aimed to examine and identify different types of bias that are relevant to ontologies, whether they are explicit or implicit. The researchers identified eight distinct types of biases present within existing production-level ontologies: (1) selectional preference; (2) gender stereotypes; (3) racial/ethnic stereotypes; (4) cultural assumptions; (5) ideological preferences; (6) political preferences; (7) religious beliefs; and (8) geographical preferences. To illustrate these biases further, examples were provided from existing production-level ontology datasets as well as literature samples such as books and articles on various topics related to each type of bias identified above.

Assessing COVID-19 Ontology Datasets

In addition to examining existing production-level ontology datasets for evidence of bias, the researchers also assessed three COVID-19 related ontology datasets that were concurrently developed during the pandemic period between March 2020 - August 2020 using their proposed framework for identifying sources of bias within an ontology dataset. The results showed that each dataset exhibited different subsets of biases found within them with varying degrees depending on its purpose and scope. For example, one dataset was found to have more gender stereotyping than another due its focus on healthcare workers who are predominantly female while another had more geographical preferences due its focus on countries affected by COVID 19 outbreaks across different continents around the world.

Conclusion & Implications

This preliminary assessment provides valuable insights into the nature of bias present within various types of existing production level as well as concurrent COVID 19 related ontological datasets which helps raise awareness about ethical considerations related to their development particularly regarding representation issues concerning information and knowledge contained therein . Furthermore , this research highlights the need for further research attention towards ethical aspects associated with developing such datasets so that potential sources or manifestations thereof can be better understood , addressed , monitored , mitigated or eliminated altogether .