Shin & Park (2019) Role of fairness, accountability, and transparency in algorithmic affordance Flashcards
Abstract
As algorithm-based services increase, social topics such as fairness, transparency, and accountability (FAT) must be addressed. This study conceptualizes such issues and examines
how they influence the use and adoption of algorithm services. In particular, we investigate how trust is related to such issues and how trust influences the user experience of algorithm services.
A multi-mixed method was used by integrating interpretive methods and surveys. The overall results show the heuristic role of fairness, accountability, and transparency, regarding their
fundamental links to trust. Despite the importance of algorithms,
no single testable definition has been observed. We reconstructed the understandings of algorithm and its affordance with user
perception, invariant properties, and contextuality. The study concludes by arguing that algorithmic affordance offers a distinctive perspective on the conceptualization of algorithmic
process. Individuals’ perceptions of FAT and how they actually perceive them are important topics for further study
What do advancements in algorithms provide?
Unprecedented venues for breakthroughs in
important decision-making fields: content curation, health and safety, security, and public management
What is F.A.T. ?
Fairness, Accountability, and Transparency
- These topics include user privacy, data policy, and ethical considerations regarding how we design and develop the algorithms, will be critical to its sustainability and long-term effects.
- This article aims to conceptualize FAT in relation to the increasing use of algorithms and
clarifies the roles of such problems in the user acceptance of algorithm services
What are the normative and operational definitions of FAT in the algorithm context? How do users perceive FAT and what constitutes FAT?
This study views an algorithm as a socially recreated artifact based on users’ cognitions and contexts
How and what does perceived FAT afford users? How is FAT related to other factors in the course of algorithm experience?
FAT is largely based on the character of algorithm service, users’ intrinsic traits, and user cognition
How does FAT influence the adoption and diffusion of algorithms? What are the roles of affordance in algorithms?
Users with higher levels of trust were observed to be more likely to see algorithms as fair, accurate, and transparent, while trust moderating the relationship between FAT and satisfaction
Fairness in algorithm contexts
Means that algorithmic decisions should not create
discriminatory or unjust consequences.
Fairness in algorithms is related to algorithmic bias
Algorithmic bias
Occurs when algorithms reflect the implicit values of the humans who are involved in coding, programming to train the algorithm
Examples of unfair discrimination:
- Banks providing loans based on race, or gender and not on financial score
- Firms hiring people based on race and not qualifications
- Realtors renting houses to specific communities and not on capability
Transparency
Involves the details of the service reasoning, and of other types of data management, involving sensible data and/or possible consequences about the knowledge that the system is gaining of the user implicitly
Accountability in algorithms
Accountability in algorithms and their application begins with the designers and developers of the system that relies on them.
Accountability often results in a loophole that goes unaddressed and uncommunicated, leaving the business vulnerable to unexpected risks for which they later maybe held responsible
Algorithmic affordances
Algorithms themselves do not have the affordance that would allow users to understand them or how best to utilize them to achieve their goals.
- Algorithmic affordances should provide users opportunities for action that people can
perceive with respect to features in their environment such as fair, transparent and
accountable
Algorithmic affordances allow a user to alter how an automated system generates an output based on data input. In this sense, algorithmic affordances characterize the range of explicit and implicit (hidden) interaction possibilities that enable the user to engage with and eventually control the algorithmic system directly and/or indirectly.
three levels of explainability of the explaianble algorithms
- Explain the intention behind how system impacts users
- Explain the data sources algorithm use and how it audits outcomes
- Explain how inputs in a model lead to outputs in a model
Important read summary for Hypotheses development!
Important read summary for Hypotheses development!