The purpose of the Institutional Review Board (IRB) is to ensure participants’ rights and welfare. As new mechanisms emerge, research compliance professionals investigate their best practices when applied in human subjects research. This blog series “Understanding Artificial Intelligence with the IRB” includes topics such as evolving research technology, artificial intelligence (AI), ethics & internet-mediated research (IMR), and future directions for comprehensive IMR research compliance. Thus, these posts offer some considerations for researchers interested in utilizing technology tools for research with human subjects purposes.
IMR has been increasingly reliable as researchers can gather data from diverse and credible online sources. Though pervasive and accessible, IMR requires specific considerations and regulations that differ from in-person research. Though in-person research includes any physical interaction during any point in research, IMR allows the recruitment of participants, communicates, and collects and analyzes data through online platforms and resources. Information may be easily accessible but online data is susceptible to hacking, false information, bot generated responses. As PIs should be cautious of where and how they collect and share, data should be stored and protected on TC-approved applications. Any new software or hardware for research must be approved by Teachers College Information Technologies (TCIT) and the IRB to ensure privacy and security standards are met. This way, researchers can prepare for the vastness and potential issues of IMR and confront the challenges of the internet that may be difficult to regulate.
In IMR, PIs should retain the consideration of knowing how to handle bots, which are AI programs that are tasked to complete objectives quickly to receive compensation from research surveys. Though there are signs of AI activity that differentiate from human responses, technology is headed in a direction that is becoming more reliant on the usage of AI. Though there are always up-to-date procedures to secure data, basic attention checks and safeguards may not be a foolproof solution to bot infiltration.
As IMR continues to develop, some researchers have turned to artificial intelligence (AI). AI aims to emulate human intelligence for different purposes, such as robotics, game playing, and machine learning (American Psychological Association [APA], 2023). AI as a digital resource is not novel, but marketing and public awareness has escalated 2022 with the publicly accessible generative bots (i.e., ChatGPT, DALL·E, Midjourney). For example, ChatGPT allows users to receive instantaneous responses in seconds and can refine them to get their desired results. Therefore, PIs may use AI in their research to quickly receive information that is usually difficult and time-consuming to obtain.
As researchers are presented with both exciting possibilities and significant ethical challenges of AI research tools, this also raises some ethical questions. Ethical concerns range from copyright infringement and privacy violations in data scraping to the complexities of the “black box” problem, where AI decision-making processes remain obscure. The emergence of generative AI also complicates issues around bias amplification and ownership. Given these considerations, researchers must use AI responsibly, ensuring they respect copyright laws, protect privacy, and maintain transparency, especially when engaging with vulnerable populations. The key is to leverage AI as a supportive tool that augments, rather than replaces, human insight and ethical considerations in research. Thus, IRB is currently developing to assist in the review process when assessing protocols using AI tools. Additionally, experts recommend having AI specialists on the IRB board for knowledge sharing and identifying risks (Blackman, 2021).
Every moment, there are discoveries and concerns with these emerging technologies that affect various organizations and industries. This blog series will offer some deeper insights into these opportunities and challenges, ethical considerations, and suggestions in properly using AI resources. For a deeper dive, please visit the other posts of this series, “Impacts of AI in Research”, “Recommendations from the Secretary’s Advisory Committee on Human Research Protections,” and “Ethical Considerations and Advice for Responsible Research in the AI Era”. We have also included a glossary of key terms that may be useful when navigating cases of AI usage in research.
Glossary of Key Terms
Artificial intelligence (AI): A subdiscipline of computer science that aims to produce programs that simulates human intelligence. There are many branches of AI, including robotics, computer vision, machine learning, game playing, and expert systems. AI has also supported research in other related areas, including cognitive science and computational linguistics (APA, 2023)
Machine Learning: A subfield of artificial intelligence, which is broadly defined as the capability of a machine to imitate intelligent human behavior. Artificial intelligence systems are used to perform complex tasks in a way that is similar to how humans solve problems. Machine learning takes the approach of letting computers learn to program themselves through experience. Machine learning starts with data, and the data is gathered and prepared to be used as training data, or the information the machine learning model will be trained on. The more data, the better the program. From there, programmers choose a machine learning model to use, supply the data, and let the computer model train itself to find patterns or make predictions (Brown, 2021). Interchangeable with Deep Learning.
Data scraping / Web scraping: An automatic method to obtain large amounts of data from websites. Web scraping plays a pivotal role in supplying data for machine learning models, furthering the advancement of AI technology. For instance, scrapped images can feed computer vision algorithms, textual data can be used for natural language processing models, and customer behavior data can enhance recommendation systems (Zyte, 2023).
Generative AI: A particular artificial intelligence (= a computer system that has some of the qualities that a human brain has, such as the ability to interpret language, recognize images, and learn from data supplied to it) that can produce text, images, etc (Cambridge Dictionary, 2023).
Large Language Model (LLM): A complex mathematical representation of language that is based on very large amounts of data and allows computers to produce language that seems similar to what a human might say. Large language models open new possibilities for text understanding and text generation in software systems (Cambridge Dictionary, 2023).