Artificial intelligence (AI) systems have the potential to revolutionize clinical practices, including improving diagnostic accuracy and surgical decision-making, while also reducing costs and manpower. However, it is important to recognize that these systems may perpetuate social inequities or demonstrate biases, such as those based on race or gender. Such biases can occur before, during, or after the development of AI models, making it critical to understand and address potential biases to enable the accurate and reliable application of AI models in clinical settings. To mitigate bias concerns during model development, we surveyed recent publications on different debiasing methods in the fields of biomedical natural language processing (NLP) or computer vision (CV). Then we discussed the methods that have been applied in the biomedical domain to address bias. We performed our literature search on PubMed, ACM digital library, and IEEE Xplore of relevant articles published between January 2018 and December 2023 using multiple combinations of keywords. We then filtered the result of 10,041 articles automatically with loose constraints, and manually inspected the abstracts of the remaining 890 articles to identify the 55 articles included in this review. Additional articles in the references are also included in this review. We discuss each method and compare its strengths and weaknesses. Finally, we review other potential methods from the general domain that could be applied to biomedicine to address bias and improve fairness.The bias of AIs in biomedicine can originate from multiple sources. Existing debiasing methods that focus on algorithms can be categorized into distributional or algorithmic.

Historically humans had various prejudices and biases like racism, classism, antisemitism, ableism, sexism, misogyny, etc. No matter which human society you live in, it would have been peppered with prejudices based on sex, gender, religion, complexion, beauty, social class, etc. in the past. However, now, we know better. The extent to which human bias can infiltrate artificial intelligence (AI) systems and cause detrimental damage is a hot topic in the tech community. To put it simply, AI bias is a problem that appears when an AI algorithm generates results that are systematically skewed due to false assumptions made during the AI training process.

As more and more companies are developing and/or utilizing artificial intelligence (AI), it is important to consider risk management and best practices to address issues like bias in AI. The National Institute of Standards and Technology (NIST) recently released a draft of its AI Risk Management Framework (Framework) and guidance to address bias in AI (Guidance). The voluntary Framework addresses risks in the design, development, use, and evaluation of AI systems. The Guidance offers considerations for trustworthy and responsible development and use of AI, notably including suggested governance processes to address bias. The Framework and Guidance will be useful for those who design, develop, use, or evaluate AI technologies.

All the sessions from Transform 2021 are available on-demand now. In early June, border officials "quietly deployed" the mobile app CBP One at the U.S.-Mexico border to "streamline the processing" of asylum seekers. While the app will reduce manual data entry and speed up the process, it also relies on controversial facial recognition technologies and stores sensitive information on asylum seekers prior to their entry to the U.S. The issue here is not the use of artificial intelligence per se, but what it means in relation to the Biden administration's pre-election promise of civil rights in technology, including AI bias and data privacy. When the Democrats took control of both House and Senate in January, onlookers were optimistic that there was an appetite for a federal privacy bill and legislation to stem bias in algorithmic decision-making systems. This is long overdue, said Ben Winters, Equal Justice Works Fellow of the Electronic Privacy Information Center (EPIC), who works on matters related to AI and the criminal justice system.

UCLA Researchers have developed a method to change the apparent race of faces in datasets that are used to train medical machine learning systems, in an attempt to redress the racial bias that many common datasets suffer from. The new technique is capable of producing photorealistic and physiologically accurate synthetic video at an average rate of 0.005 seconds per frame, and is hoped to aid the development of new diagnostics systems for remote healthcare diagnosis and monitoring – a field that has expanded greatly under COVID restrictions. The system is intended to improve the applicability of remote photoplethysmography (rPPG), a computer vision technique that evaluates facial video content to detect volumetric changes in blood supply in a non-invasive manner. Though the work, which utilizes convolutional neural networks (CNNs), incorporates previous research code published by the UK's Durham University in 2020, the new application is intended to preserve pulsatile signals in the original test data, rather than just visually changing the apparent race of the data, as the 2020 research does. The first part of the encoder-decoder system uses the Durham race transfer model, pre-trained on VGGFace2, to generate proxy target frames with the prior Caucasian-to-African component of the Durham research.

Although AI has generated excitement for the future of radiology, hopes for an automated radiological future have been dashed by reports of poor generalization of deep learning models. Models trained on images from one hospital can perform poorly when tested on images from a different one, often related to differences in disease prevalence between hospitals. Perhaps more concerning, deep learning models trained on chest radiographs (CXRs) with an underrepresentation of females have been shown to be biased for a variety of thoracic diseases; not surprisingly, these models performed better on CXRs of male patients. Biases and underrepresentation in datasets was one of several topics covered at this year's Conference on AI, Ethics, and Society, organized by the Association for the Advancement of Artificial Intelligence (AAAI) and the Association for Computing Machinery (ACM). Because AI models can reflect biases in the datasets used to develop them, detecting the presence of biases and addressing them is an important task.

The U.S. Food and Drug Administration on Thursday convened a public meeting of its Patient Engagement Advisory Committee to discuss issues regarding artificial intelligence and machine learning in medical devices. "Devices using AI and ML technology will transform healthcare delivery by increasing efficiency in key processes in the treatment of patients," said Dr. Paul Conway, PEAC chair and chair of policy and global affairs of the American Association of Kidney Patients. As Conway and others noted during the panel, AI and ML systems may have algorithmic biases and lack transparency – potentially leading, in turn, to an undermining of patient trust in devices. Medical device innovation has already ramped up in response to the COVID-19 crisis, with Center for Devices and Radiological Health Director Dr. Jeff Shuren noting that 562 medical devices have already been granted emergency use authorization by the FDA. It's imperative, said Shuren, that patients' needs be considered as part of the creation process.

Sometimes it takes something unexpected to shift people's perspectives. That's what a group of MIT and Harvard Law School researchers were aiming for when they set out to reframe fairness in AI by studying its use on the powerful rather than the powerless. They presented the results of their research in January at the ACM Conference on Fairness, Accountability and Transparency in Barcelona. In the US, over half a million people are locked up despite not yet having been convicted or sentenced--a result of pretrial detention policies. Ninety-nine percent of the jail growth since 2002 has been in the pre-trial population, much of this because of an increased reliance on bail money, according to a report by the Prison Policy Initiative.

The BBC's chart mentions a margin of error It uses data generated by artificial intelligence (AI) -- specifically, machine learning -- and it's a good example of some of the challenges that journalists are increasingly going to face as they come to deal with more and more algorithmically-generated data. Information and decisions generated by AI are qualitatively different from the sort of data you might find in an official report, but journalists may fall back on treating data as inherently factual. Here, then, are some of the ways the article dealt with that -- and what else we can do as journalists to adapt. The story draws on data from an external organisation, Ceretai, which "uses machine learning to analyse diversity in popular culture." The organisation claims to have created an algorithm which "has learned to identify the difference between male and female voices in video and provides the speaking time lengths in seconds and percentages per gender."

Is your algorithm fair and unbiased? How can you be sure that the insights it offers are correct? It's a question that's being asked with increasing frequency in the last year. That's because when it comes to machine learning, data goes into a "black box" and insights emerge on the other side. The algorithm itself is inside this so-called black box.