Filtered by themeAlgorithmic bias & fairness4 papersclear ✕

Educating Artificial Intelligence following the Child Learning Development Trajectories2024Artificial Intelligence is spreading in most daily activities. However, its develop- ment and deployment raise issues related to biases, such as gender and disability, mainly stemming from biased or incomplete datasets and lack of transparency and accountability in its algorithms. To overcome these issues, it is necessary to revert to a human-centered mindset, trying to educate algorithms rather than only train them. Adopting a human-centered approach in AI has been a first step, but it is necessary a step ahead. Indeed, recent theoretical perspectives suggest that edu- cating AI algorithms also need a profound understanding of the context of use in which it operates, adopting an approach like those in which a child is educated from birth following its developmental trajectory. By incorporating well-established educational models into the training of AI algorithms, intelligent systems based on those AI algorithms can better align with human learning trajectories, reducing bi- ases and making them more contextually aware. This paper goes in this direction, presenting an educational human-centered approach as a design methodology for artificial intelligence algorithms used within the European FRACTAL project. This proposal would pave the way toward developing more educated artificial intelligence algorithms since they are adapted to the real context of use.

UX of AI & PersonalizationResearch Papers

Interpretability, Then What? Editing Machine Learning Models to Reflect Human Knowledge and ValuesAlex Kale · 2022Machine learning (ML) interpretability techniques can reveal unde- sirable patterns in data that models exploit to make predictions— potentially causing harms once deployed. However, how to take action to address these patterns is not always clear. In a collabo- ration between ML and human-computer interaction researchers, physicians, and data scientists, we develop GAM Changer, the first interactive system to help domain experts and data scientists easily and responsibly edit Generalized Additive Models (GAMs) and fix problematic patterns. With novel interaction techniques, our tool puts interpretability into action—empowering users to analyze, val- idate, and align model behaviors with their knowledge and values. Physicians have started to use our tool to investigate and fix pneu- monia and sepsis risk prediction models, and an evaluation with 7 data scientists working in diverse domains highlights that our tool is easy to use, meets their model editing needs, and fits into their current workflows. Built with modern web technologies, our tool runs locally in users’ web browsers or computational notebooks, lowering the barrier to use. GAM Changer is available at the fol- lowing public demo link: https://interpret.ml/gam-changer. CCS CONCEPTS • Computing methodologies →Machine learning.

UX of AI & PersonalizationResearch Papers

How Designers Find Their Ways in Shaping Algorithmic SystemsJeremie Poiroux · 2022Digital products and services now commonly include algorithmic personalization or recommendation features. This has raised concerns of reduced user agency and their unequal treat- ment. Previous research hence called for increasing the participation of, among others, designers in the development of these features. To achieve this, researchers have suggested the development of better educational material and tools to enable prototyping with data and machine learning models. However, previous studies also suggest designers may find other ways to impact the development and implementation of such features, for instance through collaboration with data scientists. We build on that line of inquiry, through 19 in-depth interviews with designers working in small to large international companies to investigate how they actually intervene in shaping products includ- ing algorithmic features. We outline how designers intervene at different levels of the algorithmic systems: at a technical level, for instance by providing better input data ; at an interface or infor- mation architecture level, sometimes circumventing algorithmic discussions ; or at a organizational level, re-centering the outcome of algorithmic systems around product-centric questions. Building upon these results, we discuss how supporting designers engagement and influence on algorithmic systems may not only be a problem of technical literacy and adequate tooling. But that it may also involve a better awareness of the power of interface work, and a stronger negotiation skills and power literacy to engage in strategic discussions. Key Words: Agency, Artificial intelligence, Interventions, Machine learning, Design, User expe- rience, Algorithmic systems (2024) 33:173–204 J´er´emie Poiroux et al.

UX of AI & PersonalizationResearch Papers

Understanding perception of algorithmic decisions: Fairness, trust, and emotion in response to algorithmic managementMin Kyung Lee · 2013Algorithms increasingly make managerial decisions that people used to make. Perceptions of algorithms, regardless of the algorithms’ actual performance, can significantly influence their adoption, yet we do not fully understand how people perceive decisions made by algorithms as compared with decisions made by humans. To explore perceptions of algo- rithmic management, we conducted an online experiment using four managerial decisions that required either mechan- ical or human skills. We manipulated the decision-maker (algorithmic or human), and measured perceived fairness, trust, and emotional response. With the mechanical tasks, algorithmic and human-made decisions were perceived as equally fair and trustworthy and evoked similar emotions; however, human managers’ fairness and trustworthiness were attrib- uted to the manager’s authority, whereas algorithms’ fairness and trustworthiness were attributed to their perceived efficiency and objectivity. Human decisions evoked some positive emotion due to the possibility of social recognition, whereas algorithmic decisions generated a more mixed response – algorithms were seen as helpful tools but also possible tracking mechanisms. With the human tasks, algorithmic decisions were perceived as less fair and trustworthy and evoked more negative emotion than human decisions. Algorithms’ perceived lack of intuition and subjective judg- ment capabilities contributed to the lower fairness and trustworthiness judgments. Positive emotion from human decisions was attributed to social recognition, while negative emotion from algorithmic decisions was attributed to the dehumanizing experience of being evaluated by machines. This work reveals people’s lay concepts of algorithmic versus human decisions in a management context and suggests that task characteristics matter in understanding people’s experiences with algorithmic technologies.

UX of AI & PersonalizationResearch Papers