Computational Journalism Susume

Kazunari Fujishiro (Professor, Department of Information and Computer Science)

Not to mention "Computational Engineering," the term "Computational XX-ology (Computational ~)" is used in countless fields, including subfields of natural sciences like "Computational Physics" and "Computational Chemistry," interdisciplinary areas with computer science such as "Computational Logic" and "Computational Linguistics," and fields related to computer graphics like "Computational Geometry." In general, one can see a common characteristic of seeking new avenues through computer-based calculations. However, would it not be too limiting to confine its benefits merely to the ability to handle large amounts of data and perform high-speed approximate calculations?

I would like to point out two fundamental significances here. One is the ability to idealize the methodologies inherent to the original field, transcending physical constraints within a newly created virtual world. For example, in "Computational Photography," it is possible to reconstruct any ideal composition or shooting condition that could not have been actually captured from multiple photographs. The other is a respect for the complex thoughts and behaviors unique to humans. Did you know that "to compute" and "to consider" were originally synonymous? Computers came very close to being called "considerators."

With this in mind, this article focuses on an emerging academic field called "Computational Journalism (CJ)" and introduces some of the features of this field, which seeks to pursue the ideals of journalism by liberating users from the constraints of conventional media.

In fact, CJ aims to establish a methodology for fact-based communication with accountability, based on computer visualization technology, which has been developed to reliably acquire necessary information from various science and technology and humanities and social sciences data that are rapidly increasing and becoming more complex with the advancement of ICT ⁽¹⁾ .

Hereinafter, I would like to clarify four distinctive features of CJ through recent research examples in which I have been involved.

1) Faster-than-real-time capability: Figure 1 shows the propagation of seismic wave energy calculated by the Earth Simulator, a supercomputer that once boasted the world's fastest performance ⁽²⁾ . With this level of computational power, even if a simulation starts immediately after the epicenter is identified, it is possible to depict the propagation of the seismic waves before they actually arrive. This demonstrated the potential to visually notify the expected damage area about the degree and characteristics of the shaking in advance, simultaneously with emergency alerts.

This faster-than-real-time capability surpasses the time constraints of real-world reporting.

2) Spatio-temporal comprehensiveness: Figure 2 shows an example of tracking the temporal change in charge density distribution during a proton-hydrogen atom collision, using a manifold learning user interface that can exhaustively identify partial spacetimes with characteristic differential phase structures from time-series 3D data ⁽³⁾ .

The spatio-temporal comprehensiveness, which could not be guaranteed by conventional trial-and-error time-series analysis, resolves the problem in real-world reporting where the scope of information conveyed can vary depending on the reporter's ability, and it provides a representative way for CJ to ensure accountability.

3) Audience-driven reporting: In interactive visualization, the seamless integration of viewing and input specification reduces the user's cognitive load, allowing them to concentrate on more complex analysis tasks. Figure 3 shows an example of a preview where, after volume data representing a sheep's heart is anatomically decomposed, a user identifies a region of interest. The outer regions obscuring it are then automatically removed, and the user is navigated to a viewpoint from a direction where the geometric features are most prominent in terms of information entropy (the pie charts in the figure show the value distribution on the northern and southern hemispheres surrounding the dissected heart) ⁽⁴⁾ . The utility of this tool in computer-assisted surgery needs no explanation.

Audience-driven reporting, which prioritizes the audience's "first-person perspective" by instantly adjusting the content of the report through two-way communication between sender and receiver, demonstrates the unique flexibility of CJ.

4) Perceptually acceptable deformation: In Figure 4, the size of each object in the live scene is appropriately controlled according to its deviation from linear perspective cues (beams, railings, wood grain), and the entire screen is deformed to an extent that is unnoticeable to the audience ⁽⁵⁾ . In fact, this naturally achieves the clever composition of Uki-e (Ukiyo-e with perspective), which makes it easier to understand the scene on the forward-tilted stage while conveying the lively atmosphere of the playhouse with sufficient realism.

Non-perspective projection methods that consider perceptual tolerance show another important aspect of CJ, a return to the human element, and also suggest the potential for expansion into "emotion-based reporting" based on physiological and psychological measurements.

This article has introduced the representative features unique to Computational Journalism (CJ) as seen in recent visualization research. The faster-than-real-time capability and spatio-temporal comprehensiveness exemplify the idealization of reporting that arises from leveraging advanced information and communication media and mathematical knowledge. In addition to these, by managing visualization provenance(6), which enables the recording, tracking, and reuse of reporting procedures and content, it becomes possible to establish a processing foundation for CJ that ensures accountability. On the other hand, the two features of being audience-driven and using perceptually acceptable deformation are none other than characteristics that recognize the indispensable involvement of humans in the field of CJ.

While the examples above were all limited to one-to-one fact communication, in CJ linked to real society, it is required to establish a framework for many-to-many fact communication for the general public, such as effective reporting procedures for information disclosure and consensus building during a pandemic.

In his book " Transition of People's Way of Thinking " (1879), Yukichi Fukuzawa discussed the value of "information." As a member of the Gijuku, I would like to interpret "journalism" (報道学) as the "study" (学) of mastering the "path" (道) to "in-formation" (情報). And I firmly believe that "computation" is the modern "powerful tool" that accelerates this process.

References

(1) Cohen, S., J. T. Hamilton, and F. Turner. "Computational Journalism." *Communications of the ACM* 54, no. 10 (2011): 66–71.

(2) Fujishiro, Kazunari, Li Chen, and Yuriko Takeshima. "Large-Scale Parallel Visualization." In *Parallel Finite Element Analysis [I] Cluster Computing*, edited by Hiroshi Okuda and Kengo Nakajima, Chapter 6. Baifukan, 2004.

(3) Takahashi, S., I. Fujishiro, and M. Okada. "Applying Manifold Learning to Plotting Approximate Contour Trees." *IEEE Transactions on Visualization and Computer Graphics* 15, no. 6 (2009): 1185–92.

(4) Takahashi, S., I. Fujishiro, Y. Takeshima, and Chongke Bi. "Previewing Volume Decomposition through Optimal Viewpoints." In *Scientific Visualization: Interactions, Features, Metaphors*, edited by H. Hagen, Chapter 23. Dagstuhl Follow-Ups 2. Dagstuhl Publishing, 2011.

(5) Yoshida, K., S. Takahashi, H. Ono, I. Fujishiro, and M. Okada. "Perceptually-Guided Design of Non-Perspectives through Pictorial Depth Cues." In *Proc. CGiV2010*, 173–78, 2010.

(6) Fujishiro, Kazunari. "Collaborative Visualization." In *Fluid Informatics: The Fusion of "Fluid Dynamics" and "Information Science"*, edited by The Japan Society of Mechanical Engineers, Chapter 4. Gihodo Shuppan, 2010.