Cognitive Informatics

From Clinfowiki
Jump to: navigation, search

Cognitive informatics is the application of cognitive science to the study of informatics.[1] In addition to the field of cognitive science, cognitive informatics is informed by the related disciplines of human-computer interaction and human factors which each study human cognition and how humans interact with information systems, though with the former being more closely related to computer science and the later emerging from the field of industrial engineering. Applied to biomedicine, cognitive informatics is a framework for understanding how humans interact with health information technology (HIT) and a set of methods for designing this technology to have greater usability.

History of Cognitive Science

The field of cognitive science[2], in which cognitive informatics is rooted, is itself an interdisciplinary discipline, mixing elements of psychology, neuroscience, and computer science to study the mind and its processes. Cognitive science emerged as a discipline in the 1950's and 60's as a critique to behaviorist views of psychology which cast behavior as relationships between stimulus and response without considering internal processing. This renewed focus on cognition later became known as the cognitive revolution.

Initial research focused on the role of cognition in language understanding and production as well as the development of artificial intelligence systems to match or exceed human performance in a range of cognitive tasks. The Cognitive Science Society was founded at the University of California San Diego in 1979 and the university was a catalyst for early research in cognitive science, founding the first academic department of cognitive science in 1986.[3]

Later work continued the psychological tradition of studying cognitive processes such as memory, attention, and perception. Other areas of research which have received significant attention include the mental representation of concepts, neural systems that give rise to cognitive functions, and the neural basis of consciousness. Computational modeling of cognitive processes led to the development of artificial neural networks which form the basis of many machine learning, and deep learning methods of artificial intelligence.

Models of Cognition

Human Information Processor

The earliest and most influential systematic models of cognition cast people as "human information processors".[4] That is, not unlike a computer, human interactions with information could be understood as a series of inputs and processes which produced outputs. This led to models of cognition such as Card et al's Model Human Processor which postulated a set of resources (e.g., working memory, long term memory), processors, and actions which could be systematically mapped and timed so that the amount of time it took a human to perform a particular action (e.g., decide which of two numbers was greater) could be precisely timed. Norman's seven-stage model of human action similarly casts cognition as a systematic process of forming goals and intentions, and then selecting actions to achieve those goals.[5] This model led to the description of gulfs of execution (e.g., which action to perform) and evaluation (e.g., did my action achieve the desired intention and goal) which explain challenges with using information systems

External Cognition

Later models considered how information processing might be externalized rather than exclusively performed in the mind. For example, while a graduate student at UC San Diego, Jiajie Zhang demonstrated how varying which rules of the Tower of Hanoi problem were internally and externally represented could lead to faster and more accurate solving of the problem.[6] Kirsh et al also demonstrated how expert players of the game Tetris used tools in the game to rotate pieces to save the processing needed to mentally rotate the piece and determine if it would fit in a desired slot.[7]

Distributed Cognition

Building on these earlier models of cognition, anthropologist Ed Hutchins developed the framework of distributed cognition that holds that not only can cognition occur outside the confines of an individual mind, but that is and be distributed across tools, cultures, and time through the development and use of tools that perform certain cognitive tasks.[8] For example, the development of different map projections enable naval navigators to use straight edges to compute their location in a way that would be impossible with a globe that more accurately represented the earth's surface. The theory of distributed cognition has greatly influenced the study of HIT, particularly through the work of Vimla Patel and Jiajie Zhang.

Methods of Study

Researchers have developed numerous methods over the years to study cognitive processes related to the use of information

KLM and GOMS

Keystroke Level Modeling (KLM)[9] and GOMS (Goals, Operators, Methods, Selection Rules)[10] are related techniques for systematically modeling a human's interaction with an information system and predicting how much time completing certain actions might take. The techniques involve breaking down an activity into a set of tasks, even down to the number of keystrokes and types of decisions that need to be made at each step of the interaction. While helpful for comparing small differences in interface design, the method can be tedious to apply to longer interactions.

Cognitive Walk-through

A cognitive walk-through is technique for assessing the usability of information systems. It relies on an evaluator attempting to complete a particular task with an information system and asking a set of questions such as "What will the user's goal be at this point?" and "Will the user notice the correct action is available?" as they navigate it. This method builds on Norman's seven stage model of action to diagnose where users may have difficulty using the system. One benefit of this method is that is does not require an end-user to test the system, and can be performed even on early prototypes.

Think Aloud

Think-aloud is a method of testing system usability by having a representative participant "think aloud", that is vocalize their thoughts, while using the system to perform some task. As opposed to just observing the participant use the system, think aloud helps researchers investigate the mental models participants form about how the system works, and why they perform certain actions. One drawback of using think-aloud to study usability is that the process of thinking aloud can slow down interactions with a system, limiting the researchers ability to quantitatively measure and compare the interaction (e.g., is using system A faster than system B).

Usability Study

A usability study is a method of measuring system usability by having representative users complete a task or set of tasks using an information system with limited direction or help from the researchers. For example, a researcher may ask a participant to place a medication order in an EHR, and observe what challenges they face along the way, how long it takes them to complete the tasks, and any errors in user or system performance.

Cognitive Ethnography

Cognitive ethnography[11] is a method of understanding human cognition which is deeply rooted in the ethnographic traditions of anthropology. Cognitive ethnography employs many of the same methods as traditional ethnography (e.g., interviews, observations, co-participation in activity, artifact analysis) but with the focus of understanding what information goes where and it what form.

Computational Ethnography

Building on the method of cognitive ethnography, computational ethnography leverages the widespread availability of sensors to collect detailed information about how people interact with HIT. For example, Zheng et al describe developing and deploying a system including screen recording, multiple cameras, depth cameras, and an omni-directional microphone to study how physicians and providers interact with one another and with HIT during outpatient exams. [12]

References

  1. Patel VL, Kannampallil TG, Kaufman DR, editors. Cognitive Informatics for Biomedicine: Human Computer Interaction in Healthcare. Springer; 2015 Aug 10.
  2. https://en.wikipedia.org/wiki/Cognitive_science
  3. http://www.cogsci.ucsd.edu/about/dept-history.html
  4. Card SK. The psychology of human-computer interaction. Crc Press; 2018 May 4.
  5. Norman DA. The psychology of everyday things. Basic books; 1988.
  6. Zhang J, Norman DA. Representations in distributed cognitive tasks. Cognitive science. 1994 Jan;18(1):87-122.
  7. Kirsh D. Thinking with external representations. Ai & Society. 2010 Nov 1;25(4):441-54.
  8. Hutchins E. Cognition in the Wild. MIT press; 1995.
  9. Card SK, Moran TP, Newell A. The keystroke-level model for user performance time with interactive systems. Communications of the ACM. 1980 Jul 1;23(7):396-410.
  10. John BE, Kieras DE. The GOMS family of user interface analysis techniques: Comparison and contrast. ACM Transactions on Computer-Human Interaction (TOCHI). 1996 Dec 1;3(4):320-51.
  11. Hutchins E. Cognitive ethnography. InProceedings of the Annual Meeting of the Cognitive Science Society 2003 (Vol. 25, No. 25).
  12. Zheng K, Hanauer DA, Weibel N, Agha Z. Computational Ethnography: Automated and Unobtrusive Means for Collecting Data In Situ for Human–Computer Interaction Evaluation Studies. InCognitive Informatics for Biomedicine 2015 (pp. 111-140). Springer, Cham.

Submitted by Adam Rule