Mapping Artificial Intelligence and its Growth in Human-Centered Fields
Sadena Ahmad is a participant in the OREL undergraduate research intern program. This research is part of OREL’s Cognition Futures and Society, Ethics, and Technology theme.
What are Mind Maps?
Mind maps are a visualization tool that can be used for learning and illustrating the past and current states of understanding in a specific topical area. Although you always have to start somewhere, your research starting point should be quite different from your topical starting point (or root). A root is the source of a given topical area: an example being a key reference paper, book, or lecture. Finding the root of your mind map requires you to select several candidate starting points, then select one based on the details of your focus. During the process of creating a mind map, an individual can make and visualize connections and see the direction of growth in a specific topic.
But mind maps rarely bifurcate from a single root. Different roots lead to the same set of branches. I ran into this many times while conducting literature reviews on distinct topics. For example, Natural Language Processing (NLP) is a subfield of machine learning (stemming from deep learning) that enables computers to process text or speech data and ‘understand’ its whole meaning, including the speaker and writer’s intent (Kavlakoglu, 2020). In some examples of mind maps, NLP is factored in as a separate topic from machine and deep learning (Figure 1).
Others, however, may link it to machine learning and draw connections from there (Figure 2). Again, different origins lead to the same development. The most pertinent factor to consider is whether the links illustrated are correct and supported by references in the academic literature.
An important consideration to make is the directions for the connections — it differs depending on where you want to scope. For example, a literature review using unsupervised learning as a starting point is different in the output connections than using supervised learning as the initial starting point. Furthermore, figuring out how to expand your perspective to find the starting point means backtracking literature until you hit the dead end. Both supervised and unsupervised learning have a parallel starting point — machine learning — which can be traced back to a subfield of computer science and algorithms (Ongsulee,2017).
The most difficult part of a mind map is gathering evidence and supporting knowledge for each topic. This makes literature review a necessary step within the process to find key details in the timeline of development (with any main topic) that pinpoint a solid direction.
Mind Map creation is part of our Cognition Futures initiative.
While creating the first draft of my artificial intelligence mind map, I made several starting points that allowed me to trace the same connections to the main developments in artificial intelligence (AI — Figure 3). I also created a color code system highlighting definitions and descriptions, main ideas, floating topics, and subtopics (Figure 4).
One of the many things I’ve noticed is the similarities between my mind map and other mind maps regarding artificial intelligence once I began organizing the original ideas. This system helped me create patterns of to understand where the missing pieces to the puzzle while also pointing connections to repeating topics (Figure 5). Putting all the pieces together and diluting all the extra details, I figured out a structure that worked best for the main areas I wanted to focus on — human capacity in AI, Real-life implementation of AI like healthcare, and AI sophistication.
Healthcare and Artificial Intelligence. When looking at the growth of artificial intelligence (AI), a factor to consider is the type of field the focus is at. For example, in healthcare, the use of AI is structured differently, involving a more clinical output through disease diagnosis, health service management, predictive medicine, and clinical decision-making (Secinaro et.al, 2021). In a field like education, artificial intelligence is applied to student learning through technology such as intelligent tutoring systems (ITS). An ITS employs AI as a personalized teaching and feedback learning system without the need of a human instructor (Ahmad et al., 2021).
Although both the applications of AI in healthcare and education may fall under the same branch of optimization, routing where they are heading are different stories of development. While conducting literature reviews and exploring different areas of healthcare that have implemented AI technology, I was able to trace certain components back to main topics in my mind map. This assisted me in filtering what pieces of knowledge are most suitable for the structure.
The Gap. Innovation spreads like a contagion. It starts with one copy and continues to multiply, configure, and change its environment as time passes. Although AI innovation still has a long road to cognitive advancement before it is fully implemented into one-on-one personal settings, the roadmap for gain toward natural interaction between humans and machines is still being drawn out.
One area of focus would be the cognitive development aspects of AI programming. For example, some programs initiate an aspect of human-like capabilities, including emotion recognition and response through affective computing AI programming (D’Mello & Calvo, 2013); Brain-emotion learning inspired models (BELiMs) — a class of computational intelligence (CI) developed based on human cognition and neural structures underlying fear conditioning (Parsapoor, 2018).
Can Artificial Intelligence Be Inputted into Patient Care Settings?
Much development is still needed — especially in a field that needs intimacy, connection, and interpersonal space. AI in healthcare is still relatively new and not involved with patient care; however, early studies of AI conducted in artificial settings provide a lens into where the development is happening (Fogel & Kvedar, 2018). Promising AI studies in skin cancer screening, diabetic retinopathy, and medical adherence (Fogel & Kvedar, 2018) still vary from one-on-one patient care.
Much of the work done with integrating healthcare and AI revolves around traditional machine learning applications through precision medicine — which involves the prediction of the most successful treatment protocol for a diagnosis (Davenport & Kalakota, 2019). Other forms take a more complex approach, like deep learning in areas involving radionics and detecting clinically relevant qualities in imaging data (Davenport & Kalakota, 2019). However, the authentic components to address the fundamentals of human intelligence, such as abstraction, context, causality, explain-ability, and intelligible reasoning, are still missing from the cognitive mechanisms in deep learning AI (Singer, 2021)
A Developing Process
Creating a mind map is a learning process that requires much reading and freestyled creativity to find a comfortable baseline for connecting ideas/topics. Mind maps are an innovative tool in the assistance of illustrating an immense amount of literature review into main concepts that can sum up concluding points in the development of the leading topic. For AI, a mindmap can be a key component in constructing developmental timelines and growths — helping foresee current and future gaps in its refinement. Above all, refining topics means scoping out the most important details for the branches of a mindmap — and this can vary from one individual to another depending on their understanding of the literature.
While adding to my mind map, I could scope different concentrations of progress in the types of learning programs in AI. The dead-ends in these spectrums have allowed me to notice holes in AI growth and development to which I can turn my focus. The expectations in technological growth versus the reality of AI achievements are differentiating baselines with missing parts. Creating a mind map to illustrate this allowed visualization of the gaps in development — concluding that no matter what stage of development AI is, there is always room for new innovations and advancement in fields that may have not seen AI technology in the past.
References
Artificial Intelligence (AI) vs. Machine Learning. (2022, March 3). Columbia Engineering.
Awad, E., Dsouza, S., Kim, R. et al. (2018). The moral machine experiment. Nature, 563, 59–64.
D’Mello, S., & Calvo, R. A. (2013). Beyond the Basic Emotions. CHI’13 Extended Abstracts on Human Factors in Computing Systems on — CHI EA’13.
Davenport, T., & Kalakota, R. (2019). The potential for artificial intelligence in healthcare. Future Healthcare Journal, 6(2), 94–98.
Davenport, T., & Kalakota, R. (2019). The potential for artificial intelligence in healthcare. Future Healthcare Journal, 6(2), 94–98.
Fogel, Alexander L. & Kvedar, Joseph C. (2018). Artificial intelligence powers digital medicine. NPJ Digital Medicine, 1(1), 5.
Gjorgievska, K. (2022, April 15). Understanding Artificial Intelligence with mind mapping. iMindQ.
Gopnik, A. (2019, October 11). The ultimate learning machines. The Wall Street Journal.
Kavlakoglu, E. (2020, May 27). AI vs. machine learning vs. deep learning vs. neural networks: What’s the difference? IBM Website.
Kavlakoglu, E. (2020, May 27). AI vs. machine learning vs. deep learning vs. neural networks: what’s the difference? IBM Website.
Kumar, A. (2021, June 12). Great mind maps for learning machine learning. Data Analytics: Data, Data Science, Machine Learning, AI.
Manohar, V., & Crandall, J. W. (2014). Programming robots to express emotions: interaction paradigms, communication modalities, and context. IEEE Transactions on Human-Machine Systems, 44(3), 362–373.
Marcus, G. (2020, February 17). The next decade in AI: four steps towards robust artificial intelligence. Cornell University.
Martínez-Miranda, J., & Aldea, A. (2005). Emotions in human and artificial intelligence. Computers in Human Behavior, 21(2), 323–341.
Miller, A. (2019). The intrinsically linked future for human and Artificial Intelligence interaction. Journal of Big Data, 6(1), 38.
Ongsulee, P. (2017). Artificial intelligence, machine learning and deep learning. 15th International Conference on ICT and Knowledge Engineering (ICT&KE).
Parsapoor, M. (2018). An introduction to brain emotional learning inspired models (BELiMs) with an example of BELiMs’ applications. Artificial Intelligence Review, 52, 409–439.
Picard, R. W. (1997). Affective computing. The MIT Press.
Rule, J. S., Tenenbaum, J. B, & Piantadosi, S. T. (2020). The child as hacker. Trends in Cognitive Sciences, 24(11), 900–915.
Secinaro, S., Calandra, D., Secinaro, A. et al. (2021). The role of artificial intelligence in healthcare: a structured literature review. BMC Medical Informatics and Decision Making, 21, 125.
Secinaro, S., Calandra, D., Secinaro, A. et al. (2021). The role of artificial intelligence in healthcare: a structured literature review. BMC Medical Informatics and Decision, 21(1), 125.
Making 21, 125 https://doi.org/10.1186/s12911-021-01488-9
Singer, G. (2021, May 4). The rise of cognitive Ai. Medium. https://towardsdatascience.com/the-rise-of-cognitive-ai-a29d2b724ccc
Somers, M. (2019, March 8). Emotion AI, explained. MIT Sloan. https://mitsloan.mit.edu/ideas-made-to-matter/emotion-ai-explained
