Commentary on “Visualization in Operations Management Research”

Visualization is a visual art technique used to create graphical representations of information to support communication and exploration (Spence 2001). In the Internet of Things era, visualization is critical because it is the outcome and the validation tool of the digital foundation of complex systems and connected networks. Moreover, visualization is an important interactive platform that translates composite intelligence to enable human–artificial intelligence (AI) collaboration, which synergistically integrates AI and human intelligence for augmented decisions. The human users typically interact with visualization systems for data exploration, information integration, analytics interpretation, and insight generation. Those interpretations and insights will lead to new hypothesis generation and testing, contributing to the knowledge discovery and knowledge automation in domain sciences. As a result, the importance of the visualization has been broadly recognized by the research communities of the domain disciplines, computer vision, operations research, data science, and human factors and ergonomics, as well as industries.

The paper “Visualization in Operations Management Research” by Basole et al. (2021) lights up the pathway to the appropriate use of visualization tools in operations management (OM). This is an important research topic because it is directly related to knowledge discovery and automation when applying visualization in data-rich OM problems. The authors not only reveal how the visualization techniques augment the performance in each stage of the OM research but also point out the impacts when visualization is “used incorrectly or without sufficient consideration” (p. 1). Motivated by handling large hospital data sets, interactive visualizations were used in OM research to better interpret data sets and mitigate biased or incorrect conclusions. In their paper, the authors summarize the roles, the audience, and the challenges of the effective use of visualization in the three stages of the OM research cycle—theory/model development, theory/model testing, and translation/conveyance—which extends the traditional visualization applications in OM research. The literature is also summarized for visualization techniques; applications in different domains; augmentation in the cognitive process; and specifically, the applications in OM, such as supply chain management, network structures, and business process flows, among others.

One of the contributions of the paper is that it highlights the risks of misusing visualization, the misrepresentation by omission and misrepresentation by inclusion. In visual interface design, it is called visual mismatch negativity (Czigler 2007) (mismatch of the information and the visual representation; a lack of structure of Shneiderman’s (1996, p. 337) visual information seeking mantra, “overview first, zoom and filter, then details-on-demand”; etc.). Another contribution of the paper is that it clearly summarizes the trend of future directions in two dimensions: the availability of data for visualization and the availability of visualization techniques for the data. The authors also discuss the emerging challenges of visual rendering for privacy preserving and potential release of personal identity data.

In addition to the paper, there are three general questions for visualization: (1) What information can be provided by the visualization? (2) What insights are generated and integrated into the research problems? (3) What are the appropriate visualization and interface designs? For example, in data science research problems, visualization can be used to provide statistical information as well as domain-specific information, such as exploration of data distributions, variability, trends, variable relationships (e.g., correlation and/or causation), and the identification of novelties and anomalies. Visualization techniques are especially useful when there is a lack of deep understanding of the data distribution, structure, dimensionality, and modality; when the dimension of data sets needs to be reduced; and when one needs to determine whether additional variables are needed. Visualization can be also used to provide insights on defining the problem scope and, hence, assumptions of the proposed methods by reducing human workload in perception, cognition, and insight generation (van der Aalst 2016). Visualization should improve the effectiveness and efficiency by employing different modalities and interface designs (Bowman et al. 2002), catering to individual characteristics and the application contexts (Chen et al. 2021). In practice, many open-source visualization tools are available to facilitate the fast development of visualization designs (e.g., D3.js, Apache ECharts, Google Charts).

The paper also leads to additional future work directions in visualization. For example, it is important to improve the fidelity of the visualization for theory and model testing as a result of data uncertainty and low-dimensional representations. The testing typically evaluates how the assumptions of the theory and model are satisfied in real practice. If an assumption is violated, then further improvement or justifications should be made. However, using visualization or even mathematical or statistical methods for testing is not a trivial question. It is difficult to synthesize the validation of each assumption to the overall validity. Visualization with more advanced testing methods may provide solutions to synergistically test the theory or model. In addition, the uncertainty of the testing process will affect the fidelity of the test outcome. Visualization with statistical and optimization considerations may inform the uncertainty of the testing process itself. Moreover, the proposed theory or model cannot directly handle high-dimensional problems. Instead, it employs low-dimensional representations (e.g., a deep neural network representation for high-dimensional data) and typically comes with testing in low dimensions (e.g., statistics in a hypothesis test). New visualization techniques are needed to better project high-dimensional problems in lower dimensions, with an affordable computation workload of visualization designs.

As another example, one of the key questions for visualization in the OM research area is to create cognitive-based visualization methods to enhance the composite intelligence and decisions. This is an important pathway toward automatic hypothesis generation and testing in OM knowledge automation. In recent years, wearable physiological measures (e.g., electroencephalogram, eye movement data) have been used to improve and recommend the visualization designs (Chen et al. 2021), considering data generation (Zeng et al. 2021) and uncertainty quantification (Kang et al. 2021). The cognitive-based visualization systems will filter the right information and design the visualization interface to improve the effectiveness by predicting the cognitive status of users. The generated insights from visualization should be modeled, interpreted, and augmented with OM and AI models, such that the composite intelligence is better than either AI or human intelligence. This research will involve new theoretical advances in cognitive science, interface design, design of experiments and simulation, and knowledge modeling and automation.

Finally, it is also a promising area to explore the visualization that can provide an immersive environment for OM education. Educators can create virtual environments to interactively present education materials and engage students for active learning.