Better transparency: Introducing contextual transparency for automated decision systems

 The search tool, LinkedIn Recruiter, used by professional job recruiters to seek candidates for available job positions would be more effective if the recruiters were aware of the process used by LinkedIn to generate search query responses. This process is known as contextual transparency.

A group of scholars, headed by Mona Sloane, who is a Senior Research Scientist at NYU’s Center for Responsible AI and also a Research Assistant Professor in the Department of Technology Culture and Society, have put forth a daring new research report in the journal Nature. The term machine intelligence refers to the ability of machines or computers to perform tasks that would typically require human cognition. This includes activities such as perception, decision-making, natural language processing, and recognising patterns in complex data sets. Machine intelligence is used in a variety of fields, including finance, healthcare, manufacturing, and transportation. As advancements in artificial intelligence continue to be made, the potential applications for machine intelligence are vast and varied.

The research project involves working with two individuals: Julia Stoyanovich, who is an associate professor in computer science and engineering, data science, and the director of the Center for Responsible AI at New York University; and Ian René Solano-Kamaiko, a Ph.D. student at Cornell Tech. Aritra Dasgupta, who is an Assistant Professor specializing in Data Science, works together with Jun Yuan, a Ph.D. Candidate, both affiliated with the New Jersey Institute of Technology.

the factors and data used by the ADS to arrive at a decision, providing greater transparency for those who are affected by its decisions. This would allow individuals to have a better understanding of how decisions are made and hold the system accountable for any potential biases or errors. Essentially, the goal is to promote fairness, accountability, and ethical practice in the use of Automated Decision Systems. The ADS has certain algorithms or technological processes that are used in unique situations, which encompass both apparent and concealed factors such as the ingredients and recipe.

value of information provided by the tool—to help users understand how ADS tools operate and evaluate their results. This model includes considering factors such as user goals, the types of data used by the tool, and potential biases in the algorithm. By increasing contextual transparency, users can make more informed decisions about how to use ADS tools effectively in their specific profession or industry. A proposed model has been created by researchers for building contextual transparency in ADS tools. Different professions utilize ADS tools differently. An example of an ADS tool is LinkedIn Recruiter, which identifies candidates that meet a recruiter's criteria. The purpose of this model is to provide users with an understanding of how different factors contribute to the nutritional value of information provided by these tools. By considering factors such as user goals, data types utilized, and biases within algorithms, users can make more informed decisions when applying ADS tools in their field. Increasing contextual transparency allows for effective use of these tools across various professions and industries. The term "label" is very context-dependent. To address this, three principles of contextual transparency have been suggested as the foundation for creating transparency in each specific context, with each principle drawing on an approach from a particular academic discipline.

  • The first component (CTP 1) of this initiative focuses on the social science aspect and aims to determine how professionals use a specific ADS system and what information they require to perform their job duties more efficiently. This can be achieved through conduction of surveys or interviews with the stakeholders.
  • CTP 2: Focusing on Precision in ADS Engineering: The objective is to comprehend the technical environment of the ADS being used by stakeholders. Diverse forms of ADS work under distinct assumptions, mechanisms, and technical limitations. This principle necessitates comprehending both the data being fed into the system and its technical requirements. The process of decision-making involves using input to make a decision, and determining the manner in which the resulting output is communicated.
  • CTP 3 focuses on creating a design that is transparent in its process and specific to the desired outcomes of the ADS system. For instance, it investigates how transparency can lead to achieving the ideal outcomes, such as in recruitment where transparency can help in facilitating a more diverse pool of candidates by providing an explanation for the selection process. A model for assessing or assigning a position or level of importance to something based on certain criteria is referred to as a ranking model.

In a study focused on the use of LinkedIn Recruiter by recruiters who use Boolean searches (queries using terms like AND, OR and NOT) to obtain ranked results, researchers explored the potential benefits of context transparency. Their findings revealed that recruiters usually verify and cross-check rankings derived from ADS rather than blindly trusting them. Frequently, in order to achieve more accurate results, adjustments to keywords are made when ranking outputs. Researchers were informed by recruiters that the absence of transparency in ADS presents difficulties in recruiting for diversity.

In order to meet recruiters' transparency requirements, experts propose including both passive and active factors in the nutritional label of contextual transparency. Passive factors refer to information that pertains to the overall functionality of the system, as well as the recruiting profession. Overall, active components include details that are unique to the Boolean search query and can vary.

enhance the precision of subsequent searches. The nutrient information sheet will be integrated into the regular LinkedIn Recruiter process, offering users valuable insights to evaluate how well their search results meet their intended purpose and allowing them to modify their search criteria to narrow down their queries more accurately. Produce improved outcomes.

available). To determine if the ADS transparency intervention was effective in bringing about expected changes, experts propose conducting interviews with stakeholders to assess any changes in their attitudes towards and use of ADS, as well as having participants keep diaries documenting their professional practices. Additionally, researchers suggest performing A/B testing if possible. Reworded version: It is advisable to take all necessary precautions in order to avoid any potential risks that may arise. In doing so, one can minimize their chances of encountering any adverse outcomes or negative consequences. By taking the necessary steps and being vigilant, one can ensure their safety and well-being to the best of their ability.

The method of contextual transparency can be employed to meet the demands of AI transparency outlined in emerging US and European regulations, including NYC Local Law 144 of 2021 and the EU AI Act.

Post a Comment

Previous Post Next Post