Fairness issues in AI systems that augment sensory abilities

Findlater, Leah; Goodman, Steven; Zhao, Yuhang; Azenkot, Shiri; Hanley, Margot

doi:10.1145/3386296.3386304

Cited by 17 publications

(9 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, the intention to provide 'equal' access to information raises the question of fairness in the context of assistive technologies. Prior works discussed AI-related fairness challenges in the context of accessibility and the ethical implications to decide what information AI should provide to users [34,42,55,97]. Findlater et al focused on balancing the privacy concerns of the primary users and others raised by sensory augmentation.…”

Section: Fairness and Equity In Information Accessmentioning

confidence: 99%

“…Hamidi et al reported how identifying gender through facial features can challenge the autonomy of transgender people and reinforce gender binary attitudes [67]. Furthermore, their use can adversely afect marginalized populations (such as people of color and transgender people) [42,55,77,105,125] by increasing inequality [102]. In the context of algorithmic errors, prior work has also highlighted how facial recognition systems are incorporated within systems of policing and surveillance where misclassiications can create signiicant problems (e.g., if the algorithm wrongfully identiies someone as a suspect) [14,68].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Shared Privacy Concerns of the Visually Impaired and Sighted Bystanders with Camera-Based Assistive Technologies

Akter

Ahmed

Kapadia

et al. 2022

ACM Trans. Access. Comput.

View full text Add to dashboard Cite

Camera based assistive technologies can provide people with visual impairments (PVIs) visually derived information about people in their vicinity. Furthermore, the advent of smart glasses offers the possibility of not only analyzing visual information in front of the wearer, but also behind them through an extended field of view. Although such ‘visually available’ information can enhance one’s social interactions, the privacy and ethical implications for automated judgments about bystanders especially from the perspective of PVIs remains underexplored. To study the concerns of both bystanders and PVIs with such technologies, we conducted two online surveys with visually impaired participants as wearers (N=128) and sighted participants as bystanders (N=136). Although PVIs found some types of information to be improper or impolite (such as someone’s weight), our overarching finding is the shared ethical concern between PVIs and bystanders related to the fallibility of AI, where bystanders can be misrepresented (algorithmically) by the devices. These mischaracterizations can range from occasional, unexpected algorithmic errors (e.g., errors in facial recognition) to the questionable use of AI for determining subjective, social constructs (such as gender). Based on our findings, we discuss the design implications and directions for future work in the development of camera based assistive technologies while mitigating the ethical concerns of PVIs and bystanders.

show abstract

Section: Fairness and Equity In Information Accessmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Shared Privacy Concerns of the Visually Impaired and Sighted Bystanders with Camera-Based Assistive Technologies

Akter

Ahmed

Kapadia

et al. 2022

ACM Trans. Access. Comput.

View full text Add to dashboard Cite

show abstract

“…Machine learning is a type of AI that provides machines with the ability to learn without being explicitly programmed [8]. Machine learning technology leverages advances [9] in computer vision, speech recognition, and auditory scene analysis to sense, interpret, and offer new ways to help people with disabilities. The taxonomy of the machine learning accessibility tools for people with disabilities (visually, hearing, and physically impaired) is illustrated in Figure 2.…”

Section: Machine Learning For People With Disabilitiesmentioning

confidence: 99%

An Overview of Machine Learning and 5G for People with Disabilities

Domingo¹

2021

Sensors

View full text Add to dashboard Cite

Currently, over a billion people, including children (or about 15% of the world’s population), are estimated to be living with disability, and this figure is going to increase to beyond two billion by 2050. People with disabilities generally experience poorer levels of health, fewer achievements in education, fewer economic opportunities, and higher rates of poverty. Artificial intelligence and 5G can make major contributions towards the assistance of people with disabilities, so they can achieve a good quality of life. In this paper, an overview of machine learning and 5G for people with disabilities is provided. For this purpose, the proposed 5G network slicing architecture for disabled people is introduced. Different application scenarios and their main benefits are considered to illustrate the interaction of machine learning and 5G. Critical challenges have been identified and addressed.

show abstract

“…In the feld of accessibility, human-centered ML applications can allow disabled users to personalize data-driven assistive technology to meet their individual needs [40]. However, training an ML-enabled application as a personal assistive technology can itself be inaccessible when it requires skills and abilities similar to those the application is intended to support [23,40]. For example, a blind or visually impaired user is likely unable to use visual feedback when capturing images for personalizing an object recognizer-a challenge that Kacorri et al and others (e.g., [41,69]) frst examined via studies of users' needs in this context, and more recently began addressing through active feedback techniques to assist in image capture [49].…”

Section: Human-centered Machine Learningmentioning

confidence: 99%

Toward User-Driven Sound Recognizer Personalization with People Who Are d/Deaf or Hard of Hearing

Goodman

Liu

Jain

et al. 2021

Proc. ACM Interact. Mob. Wearable Ubiquitous Technol.

Self Cite

View full text Add to dashboard Cite

Automated sound recognition tools can be a useful complement to d/Deaf and hard of hearing (DHH) people's typical communication and environmental awareness strategies. Pre-trained sound recognition models, however, may not meet the diverse needs of individual DHH users. While approaches from human-centered machine learning can enable non-expert users to build their own automated systems, end-user ML solutions that augment human sensory abilities present a unique challenge for users who have sensory disabilities: how can a DHH user, who has difficulty hearing a sound themselves, effectively record samples to train an ML system to recognize that sound? To better understand how DHH users can drive personalization of their own assistive sound recognition tools, we conducted a three-part study with 14 DHH participants: (1) an initial interview and demo of a personalizable sound recognizer, (2) a week-long field study of in situ recording, and (3) a follow-up interview and ideation session. Our results highlight a positive subjective experience when recording and interpreting training data in situ, but we uncover several key pitfalls unique to DHH users---such as inhibited judgement of representative samples due to limited audiological experience. We share implications of these results for the design of recording interfaces and human-the-the-loop systems that can support DHH users to build sound recognizers for their personal needs.

show abstract

Fairness issues in AI systems that augment sensory abilities

Cited by 17 publications

References 18 publications

Shared Privacy Concerns of the Visually Impaired and Sighted Bystanders with Camera-Based Assistive Technologies

Shared Privacy Concerns of the Visually Impaired and Sighted Bystanders with Camera-Based Assistive Technologies

An Overview of Machine Learning and 5G for People with Disabilities

Toward User-Driven Sound Recognizer Personalization with People Who Are d/Deaf or Hard of Hearing

Contact Info

Product

Resources

About