Future of AI in Libraries
Published: March 15, 2021 by Dr. Souvick Ghosh
Artificial Intelligence (AI) has been one of the most transformative technologies since electricity (Ng, 2018). Electricity transformed the world as we see it, making innovations possible in other industries like healthcare, manufacturing, and transportation. Similarly, AI and Machine Learning (ML) solutions have found applications in finance, education, healthcare, and manufacturing. For simplicity, I will use AI to represent all developments in artificial intelligence, machine learning, and deep neural networks (also known as deep learning). Cheaper storage and social media have both acted as catalysts for large-scale data collection. To process this data using human labor would be prohibitive in terms of cost and time. Developments in AI make it possible to analyze these enormous amounts of data (“Big data”) for pattern recognition. For example, the healthcare domain uses patient electronic health records (EHR) for disease diagnosis, prognosis, and treatment. Financial analysts use the market data for risk assessment and investment decisions. Educators use student data to track progress and identify weaknesses. Product marketing and recommendation have become dependent on AI too. As the value of product usage data keeps growing, end users are no longer the customers of the service. Users have become the product themselves as the data they generate drive multiple industries forward using the power of AI.
AI is transformative and yet has the potential to be misused. Targeted marketing violates the privacy of the user. AI-powered information systems have also been criticized for being biased and racist (Noble, 2019). Microsoft’s chatbot Tay was racist and offensive (Lee, n.d.), while Amazon’s Alexa will nudge you toward a product or subscription sold by Amazon. As information becomes abundant, quality control has become a significant issue. Libraries have traditionally been the gatekeepers of information. With AI taking control and libraries reluctant to evolve and change, misinformation has been rampant. Users rely on voice-assistants (Apple’s Siri, Amazon’s Alexa, Google’s Assistant) to get answers to simple questions. As these proprietary systems are developed and maintained by large technology giants, and the AI models used are black boxes, the user is left with limited choice.
In my research and teaching, I work toward a human-centered application of AI. The awareness of how machine learning and deep neural networks work could remediate some of the concerns around AI. Through the development of open-source systems that are explainable, transparent, and fair, we could bring the focus back on the user. My research is around the development of conversational systems, which include both chatbots and voice-based personal assistants. Such systems are very good at answering simple questions, but unlike librarians, they cannot evaluate the nature of the user’s information need. In my research (Ghosh, 2020; Ghosh & Ghosh, 2021), I have explored strategies to improve the natural language understanding of conversational search systems. I have also looked at how to facilitate the conversation between the user and the system. The advantage of voice-based systems is the hands-free accessibility, which helps special needs individuals and is also situationally beneficial. When typing is erroneous or risky, the user can speak to the system (Trippas et al., 2018; Turunen et al., 2012). First responders can be aided and directed during emergencies to avoid being bottlenecked by the finite number of dispatchers (Madeiros, n.d.). Also, experts in many fields are not necessarily adept at using a graphical user interface. A conversational system allows these experts to verbally access the search system with a priority to decision making without any intimate knowledge of how to use the specific search system (Harborne et al., 2018). Students who have had to switch between many different learning management systems (Canvas, Moodle, Blackboard) know how nuanced differences between very similar systems can be.
The libraries need to evolve and embrace the AI revolution to make AI and conversational systems more accessible and human-centered. Chatbots are already being used in several websites that handle questions from patrons, offer directional advice, and point patrons to relevant resources. Deploying conversational systems in libraries and museums will help patrons with their queries and increase accessibility. Additionally, machine learning could be used to gather insights about the usage statistics. Which books are loaned at which time of the year? How do patrons search for information? Machine learning can assign subject headings to full texts, index and classify documents, and develop vocabularies. Reference managers like Mendele and ResearchRabbit use machine learning to identify and recommend research papers based on the user’s reading habits. In one of my projects with OCLC research, we used clustering techniques to automatically identify publisher entities using MARC records. But the relationship between libraries and AI goes beyond the application of AI for library functions. Libraries can offer AI education to the unemployed, the underemployed, and the homeless (Johnson, n.d.). Also, libraries could help create those community spaces where the patrons can interact with information safely and privately. AI is here to stay, so the sooner the libraries include AI in their operations, the better.
Ghosh, S. (2020). Exploring intelligent functionalities of spoken conversational search systems. Rutgers University-School of Graduate Studies.
Ghosh, S., & Ghosh, S. (2021). Classifying Speech Acts using Multi-channel Deep Attention Network for Task-oriented Conversational Search Agents. Proceedings of the 2021 Conference on Human Information Interaction and Retrieval, 267–272.
Harborne, D., Braines, D., Preece, A., & Rzepka, R. (2018). Conversational control interface to facilitate situational understanding in a city surveillance setting.
Johnson, B. (n.d.). FEATURE – Libraries in the Age of Artificial Intelligence. Retrieved March 13, 2021, from https://www.infotoday.com/cilmag/jan18/Johnson–Libraries-in-the-Age-of-Artificial-Intelligence.shtml
Lee, P. (n.d.). Learning from Tay’s introduction – The Official Microsoft Blog. Retrieved March 11, 2021, from https://blogs.microsoft.com/blog/2016/03/25/learning-tays-introduction/
Madeiros, J. (n.d.). Alan: The AI Companion For First Responders. Retrieved October 14, 2020, from https://www.voicesummit.ai/blog/alan-the-ai-companion-for-first-responders
Ng, A. (2018). AI is the new electricity. O’Reilly Media.
Noble, S. U. (2019). Algorithms of Oppression. In Algorithms of Oppression. https://doi.org/10.2307/j.ctt1pwt9w5
Trippas, J. R., Spina, D., Cavedon, L., Joho, H., & Sanderson, M. (2018). Informing the Design of Spoken Conversational Search: Perspective Paper. Proceedings of the 2018 Conference on Human Information Interaction&Retrieval, 32–41.
Turunen, M., Hakulinen, J., Rajput, N., & Nanavati, A. A. (2012). Evaluation of mobile and pervasive speech applications. Speech in Mobile and Pervasive Environments, 219–262. https://doi.org/10.1002/9781119961710.ch8
Post new comment