The multistability of predictive technology in nuclear disasters

Analytical perspectives

SPEEDI’s designers: Prediction as supporting advice

SPEEDI was derived from the strong political purpose of social control. Prompted by the US Three Mile Island (TMI) accident in 1979, Japan’s Nuclear Safety Commission (NSC) established a special expert group and published a technical guideline for emergency management around nuclear power plants in June 1980 (Nuclear Safety Commission [NSC], 1980). This guideline stipulates that ‘if a large release of radioactive materials happens, or is about to happen, the impact of abnormal situations should be reduced as much as possible by quelling residents’ unrest and controlling disruption considering the peculiarity and similarity of nuclear disaster’ (NSC, 1980, p. 3). It also contains the following statements, which may be regarded as an archetype of the ‘deficit model’ (Wynne, 1991) through the STS scholarly lens: ‘the knowledge of nuclear emergency preparedness must be distributed, and the general public must be enlightened, for residents to act orderly and follow the instructions from the local headquarters for disaster control, without unrest and disruption’ (NSC, 1980, p. 6). This approach mirrors that of the (U.S.) President’s Commission on the Accident at Three Mile Island (1979), which stated that ‘the actual release will have a negligible effect on the physical health of individuals’, ‘the major health effect of the accident was found to be mental stress’ caused by the management failure of the regulatory agency and relevant organizations, as well as the fact that ‘some significant fraction of the population in the immediate vicinity voluntarily left the region’ (p. 13, emphasis added). Under this social control framework, it should not be each resident but a nuclear technocracy that decides evacuation. SPEEDI was developed to make this technocratic decision.

In the guidelines, emergency decisions of protective actions were based on the projected dose. For instance, if the external exposure dose is expected to exceed 10 rem in a particular area near nuclear facilities, evacuation in this area will be mandatory. If it is likely within 5-10 rem, people will be mandated to shelter in place. In the early 1980s, the projected dose was manually calculated using an approximation formula for atmospheric dispersion and acrylic plates copying the maps surrounding the nuclear power plant. To improve the reliability of dose projections, the nuclear safety research plan of the fiscal year 1980 incorporated emergency environmental radioactivity studies, including the development of a predictive simulation system. In response to this, experts from the Japan Atomic Energy Research Institute led the development of SPEEDI, based on Atmospheric Release Advisory Capability (ARAC), a system developed by the US (Imai et al., 1985). Eventually, the NSC officially institutionalized the use of the SPEEDI by revising the technical guidelines of nuclear emergency management in 1992. Stipulated as ‘basic information for deciding the areas of environmental monitoring and protective actions including evacuation’ (NSC, 1992, p. 50), SPEEDI’s prediction maps visualizing the distribution of radioactive materials and projected dose were considered essential for emergency decisions. Understandably, the shift from ‘manual’ to ‘simulated’ predictions may lend ‘an aura of objectivity’ (Ballestero, 2014; Neale & May, 2020).

While the guidelines for nuclear emergency management have been frequently revised, the basic concept of social control for ‘quelling residents’ unrest and controlling disruption’ has been maintained even in the last revision in August 2010 (NSC, 2010). In this framework, local people were considered subjects who follow government instructions, not autonomous agents who think and behave actively. SPEEDI’s history shows that its prediction maps had been developed and mobilized to ensure social control by constructing a public imaginary of epistemic and psychological certainty under nuclear emergencies.

The SPEEDI’s maps were not assumed to be directly available to the public. The designers meant for the users of this system to be the decision-makers and experts who had expertise similar to themselves—‘environmental experts’ in contrast to ‘reactor experts.’ One of the designers emphasized in an interview the reliability of their models because they were ‘scientifically’ validated through peer reviews of academic journals. This contrast with similar calculation models, such as the Dose Information Analysis at Nuclear Accident (DIANA) of the Tokyo Electric Power Company, that are owned by the industry and have not undergone scholarly examinations. Thus, it is understandable that ‘environmental experts’, to a large extent, are proud and confident of the reliability of SPEEDI predictions, particularly in terms of local weather forecasts and atmospheric dispersion modelling. However, in interviews, they also identified non-negligible uncertainties, including those in the source term information and local topographic effects. They thought it would be impossible to determine evacuation areas based only on the absolute value of the projected dose provided by SPEEDI (interviews, senior researchers of atmospheric dispersion at a public institution). Instead, experienced experts should interpret the prediction maps appropriately, and decision-makers should be carefully informed. The designers’ pre-spectival focus was not simply on the SPEEDI maps but the expert advice connecting the predictive simulation to emergency decisions. In their sense, the future consequences of nuclear disaster could emerge as a gestalt only through a comprehensive analysis of various information, including SPEEDI predictions and radiation monitoring results and their prudent interpretation. They assume that it requires well-developed interpretive skills and expertise, notably in atmospheric dispersion and environmental radiation. This is why they thought that SPEEDI maps should not be directly disclosed to laypeople who do not have sufficient expertise. It may reflect their form of life as ‘environmental experts’ involving their sense of dealing with simulation and its uncertainties.

Using ANT terms, there has been a solid social control network in Japanese nuclear disaster management from its dawn. However, directly forcing residents to evacuate was not initially delegated to this prediction technology. Instead, it was to support expert advice for rendering emergency decisions more reliable and authoritative and thus to make residents follow the governmental instructions.

Pre-disaster: Prediction as a tool for evacuation drills

Fortunately, nuclear disasters are rare, so there are few opportunities to experience severe accidents and massive evacuations. Thus images of nuclear disasters were constructed not by first-hand experiences of radiological emergencies but primarily through training drills. Before the Fukushima accident, a nuclear emergency accompanied by residents’ evacuation occurred only once in Japan: the JCO criticality accident in Tokai village on September 30, 1999. Consequently, for many practitioners in nuclear hosting areas, only the nuclear disaster drills had established the collective images of nuclear emergencies and practical ways of taking advantage of SPEEDI.

As shown in Figure 1, in past nuclear emergency drills, the decision-making process of evacuation directly referred to the prediction maps. The maps provided by SPEEDI contain several contour lines representing the projected radiation exposure dose. A team in charge of residents in the local headquarters of nuclear emergency management, composed mainly of local government officials, makes a draft plan of protective action areas, and then of protective action districts. Interestingly, this process establishes a boundary between technoscience and society: the protective action areas are generated geometrically, almost mechanically, from the expected shape of the radiation plume, but the plan for protective action districts considers ‘societal factors’ including logistics, populations, and residents’ emotions (Secretariat of NSC [Nuclear Safety Commission], 2006). Only the second step, the production of protective action districts, required expertise and judgment. The uncertainties associated with SPEEDI calculations themselves, such as weather forecasts and the projected direction of dispersion dependent on the time of massive release, were considered to be adequately addressed by the margin the protective action areas provide around the ellipsoid radiation contour. Instead, the practitioners’ pre-spectival focus was directed to translating the scientifically provided prediction maps into socially workable evacuation zoning, reflecting the actual conditions of local communities. In their sense, while the SPEEDI maps were construed as solid scientific evidence that could almost prescribe evacuation zoning, addressing societal uncertainties was deemed more important than the techno-scientific uncertainties inherent to predictive simulations.

OPEN IN VIEWER

Figure 1.

Procedural flow of evacuation zoning in past nuclear emergency drills.

In the shared practice established through the nuclear emergency drills, residents were considered subjects required to obey government instructions. Those who lived in the communities subject to the disaster drills had to take evacuation actions without getting flustered, including coming to a specified place at a designated time and boarding pre-arranged buses calmly. Under the ANT account, SPEEDI was enrolled in a social control network similar to the designers’ one. However, as pre-spectival focus shifted away from the expert advice to localizing prediction maps, the need for comprehensive analysis and prudent interpretation receded into the background. In conjunction with this shift, the delegated role of SPEEDI became more direct and more significant in this network: from providing the supporting information for expert advice to providing scientific evidence that almost directly prescribes evacuation areas.

Post-disaster: Prediction as self-protection

While there were controversies around the opportunities for using SPEEDI after the Fukushima disaster, the dominant stable use of this technology was publicly established during these ten years, as seen in the exhibition of the Memorial Museum. Representing this stability, the ICANPS report states, ‘if unit quantity emission predictions had been provided, the various municipal bodies and residents would probably have been able to decide which evacuation routes and directions were most suitable for them, taking advantage of their first-hand knowledge of local road conditions’ (ICANPS, 2011, p. 579). Note that ICANPS assumes that the users of this technology are local governments and residents rather than the central government or atmospheric dispersion experts. This post-accident dominant stability differs from the designers’ intention and past customary practices. As mentioned above, because of their pre-spectival focus the SPEEDI designers did not envision directly disclosing simulation results. In contrast, in the dominant post-accident discourse, local government officials and residents should have been informed of SPEEDI maps so they could have improvised the directions and routes of evacuation. In this stability, prediction maps are regarded as neutral media representing situations several hours into the future, and these may be easily understood without special scientific expertise. The pre-spectival focus is concentrated on the prediction maps, while other parts of the making-future-present process, such as the uncertainties of simulative calculation and their careful interpretation, are relegated into the background. In postphenomenological terms, those processes are ‘transparent in use’ (Ihde, 2009), when the user’s attention is directed toward not the tool but the world, such as in the case of the blind person’s cane (Merleau-Ponty, 1962). Such pre-spectival focus, in conjunction with the thirst for information in the acute phase of the Fukushima disaster, comes to link with a new actor-network, the ‘self-determination network’.

The self-determination network gradually emerged in the lessons-learning process of this accident, and it was consolidated through the controversy regarding the restarting of other nuclear power plants. At the initial stage of the Fukushima accident, the Japanese government issued instructions for protective actions, including evacuation and sheltering, six times from the night of March 11 to the morning of March 15, 2011. ¹ However, due to damaged communication infrastructure and incomplete messaging, many residents received evacuation instructions only via TV news (Independent Investigation Commission on the Fukushima Nuclear Accident, 2014), and approximately 70% of people living in five nearby towns were obligated to change their location more than four times. Such hardships led people to the idea that they should protect themselves, because the government is undependable at critical moments. Indeed, the NAIIC’s survey shows that many Fukushima evacuees regard more proactive use of SPEEDI as desirable. Notably, approximately half of the residents in Namie town, located northwest of the Fukushima-Daiichi nuclear power plant and heavily affected by radioactive plumes, think that if the prediction maps had been disclosed to them, they could have avoided unnecessary radiation exposure (National Diet of Japan Fukushima Nuclear Accident Independent Investigation Commission [NAIIC], 2012).

The use of SPEEDI became a hot issue in other areas hosting nuclear power plants. In the process of revising the guidelines for nuclear emergency preparedness, local stakeholders pushed for increasing the reliability of prediction and its proactive use in evacuation decisions (Zengenkyo, 2012). Moreover, to restart the nuclear power plants that stopped their operations after the accident, attention was paid to the effectiveness of evacuation plans, in addition to compliance with newly established regulatory requirements. Although the NRA had decided to abandon SPEEDI in October 2014 (as described in the next section), some prefectures and towns sought the possible use of SPEEDI prediction in their disaster management. For example, Nagaoka city, located near the Kashiwazaki-Kariwa nuclear power plant, which has the same operator, TEPCO, as Fukushima Daiichi, organized a study meeting with neighboring cities to discuss nuclear safety and disaster management, including an exploration of prediction-informed evacuations since September 2011 (Nagaoka City, 2011). Similarly, Niigata prefecture established a special committee for independently examining evacuation measures, including ways of using systems of predicting radionuclide dispersion (Niigata Prefecture, 2017). The National Governors’ Association recommendations to the national government included the proactive use of SPEEDI predictions (National Governors’ Association, 2014). While local governments do not have legal authority over nuclear safety in Japan, the governors have non-negligible political power around nuclear issues because the governors’ consent is deemed de facto necessary in restarting the nuclear facilities after an incident (Sugawara & Shiroyama, 2010). In response to this, the Ministerial Council for Nuclear Power Utilization made a political decision: ‘[T]he national government shall never prevent the prefectural/municipal governments from referring to SPEEDI calculation results at their own responsibility, both for their nuclear emergency decision-making and emergency drills’ (Cabinet Secretariat of Japan, 2016). The Prefectural Government Association on Nuclear Power (PGA), composed of the governors of prefectures where nuclear facilities are located, has repeatedly petitioned the government to incorporate ‘information that predicts the dispersion of radioactive materials’ into nuclear emergency management (Prefectural Government Association on Nuclear Power [PGA], 2020).

Such positive views of SPEEDI use are shared broadly among local stakeholders in nuclear hosting areas. A resident said that the nondisclosure of SPEEDI was outdated, pointing to how paternalistic doctors in the past hesitated to show the medical tomography scans to patients. In the self-determination network, residents are autonomous individuals who can actively interpret information and decide evacuation actions. Thus, evacuation is enacted as a voluntary action by residents, rather than a controlled measure following governmental instructions. From the viewpoint of this new network, nondisclosure of prediction maps can result from ‘elite panic’ (Solnit, 2010) in the central government and among experts, which may hinder the self-determination of the local public.

Nuclear Regulation Authority: Prediction as a source of misunderstanding

Meanwhile, some experts in nuclear safety and regulatory authority expressed concerns regarding the post-accident dominant stability of SPEEDI. In fact, NRA eventually regarded SPEEDI as dangerous: Following their decision on the non-use of SPEEDI for emergency decisions on October 2014, they cut off the SPEEDI budget, took down the designated terminal devices from local government offices, and completely banned its use on March 2016.

The NRA experts who assume leadership in nuclear emergency management come from Level 3 probabilistic risk assessment (PRA) and severe accident studies. Level 3 PRA is a methodology for probabilistically estimating the radiological consequences of nuclear accidents based on Level 1 and Level 2 PRA results. The former estimates the pathway of core melt accident and its frequency. The latter evaluates the possible damage to the primary containment vessel after core melt and consequent radioactive release (source term and frequency). Although Level 3 PRA addresses the issue of health and environmental impacts outside the facilities, its assessment is deeply connected to what happens inside the nuclear reactor. While the SPEEDI developers are apt to see the nuclear disaster from the side of offsite and environment, the PRA and severe accident experts tends to regard it as an extension from the on-site and reactor accident. In their professional sense of risk assessment and severe accidents, the ‘environmental experts’ do not adequately understand what will happen inside the nuclear reactor and how a nuclear accident will progress (interview, a senior ‘reactor expert’ at a public institution). The ‘reactor experts’ keenly realize the difficulty—almost impossibility—of predicting the behaviors of a nuclear reactor in a severe accident, wherein highly complex physical and chemical reactions can occur under very high temperatures and pressure, and the significant uncertainties when estimating source term information through a Level 2 PRA, including those of decontamination factors, chemical compositions, and release channels. Additionally, the PRA experts emphasized the importance of squarely addressing the uncertainties of risk assessment by representing the assessors’ degree of belief as a probability distribution (Wu et al., 1990). From their point of view, SPEEDI’s prediction maps do not acceptably incorporate or express uncertainty. First, the source term information, which is essential for SPEEDI’s simulation, cannot be estimated with a specific degree of reliability because of the difficulty in predicting the accident’s progress. Second, the methods of visualizing prediction maps do not represent critical uncertainties, including source term information, weather forecasts, and deposition rates on the soil. Thus, they vehemently denied both the post-Fukushima dominant stable use and the past customary use of the SPEEDI. One of the NRA experts said that it might not be entirely impossible to use SPEEDI as a source of supplementary information if the maps are prudently interpreted by qualified experts having enough knowledge of reactors and severe accidents—not just scientifically-minded ‘environmental experts’. He added with resignation that it would be difficult to establish such an emergency framework in reality (interview, a senior expert of Level 3 PRA and nuclear emergency management at a public institution). The reactor experts’ pre-spectival focus is mainly on the stage before atmospheric dispersion. Reflecting the ‘form of life’ nurtured in their professional experiences of PRA, it is too dangerous to straightforwardly use atmospheric dispersion maps without a cautious recognition of the difficulty of predicting the trajectories of an accident. For them, SPEEDI is construed as a source of misunderstanding unless its uncertainties are carefully addressed. In a very different context, a fire behavior analyst says: ‘maps often communicate the “wrong things” at the most important time simply through their formal properties’ (Neale & May, 2020, p. 851).

In February 2013, the NRA abolished the NSC’s guidelines ² and formulated new guidelines for nuclear emergency management. It introduced a novel approach wherein protective actions are mechanically taken, triggered by observable parameters, such as the duration of safety functions’ loss and the drawdown of the water level in the spent fuel pool. Instead of using the Japan-specific frameworks, heavily dependent on predictive simulations, to optimize evacuation zones, the new framework, following international practices, adopts concentric-shaped zoning regardless of wind direction. Usually, the area within a 5 km radius of a nuclear power plant is designated as the precautionary action zone (PAZ), wherein residents must evacuate precautionarily before a significant release occurs. From 5 to 30 km is an urgent protective action planning zone (UPZ), wherein the necessity of evacuation is decided based on the observable value of the air dose rate after releasing radioactive materials. After the release of radioactive materials, if the measured values of the air dose rate exceed the pre-set criteria (for instance, 500 μSv/h), people in that district are required to evacuate (Nuclear Regulation Authority [NRA], 2019). In this new framework, prediction plays no role: Observable parameters automatically trigger the necessary course of action without requiring estimations and judgments from experts or politicians (Sugawara, 2021; Sugawara & Juraku, 2018).

In contrast to the post-accident dominant stability, this new approach depicts residents needing to wait for governmental instructions. People in the UPZ are required not to evacuate when evacuation in the PAZ is ongoing, to prevent traffic from dangerously slowing the evacuation. Nuclear experts think that stopping people in UPZ from evacuating before a significant release is one of the most crucial challenges (interview, a senior expert of Level 3 PRA and nuclear emergency management at a public institution). When calculating the evacuation time estimation, those who decide to evacuate themselves are termed ‘shadow evacuees’ (Shimada & Takahara, 2021), which connotes a deviation from proper behavior. The NRA’s views on evacuation seem to reconsolidate it into the conventional social control network, in competition with the new self-determination network—though without enrolling SPEEDI.

Rivalry between two actor-networks

Table 1 provides an overview of the multistable uses of SPEEDI. It suggests that how the relevant actors perceive the predictions of a nuclear disaster in each stability depends on the direction of their pre-spectival focus, which is related to their form of life. Interestingly, as wider stakeholders get involved in nuclear disaster management, the pre-spectival focus on the making-future-present process narrows. At the development stage of SPEEDI, the designers who recognized the uncertainties and limitations of their simulations construed the future of nuclear disaster not as simply represented on a prediction map itself but as emerging as a perceptual gestalt through consolidation of heterogenous information and well-trained interpretive frameworks. They thought that only environmental experts like themselves would have such hermeneutic skills, so SPEEDI should not be disclosed directly to the public. However, after SPEEDI has been institutionalized and used in nuclear emergency drills, the recognition of the need for prudently interpreting the uncertainties and consolidating various information receded into the background. The practitioners’ pre-spectival focus was on the process of adjusting scientifically predicted maps into socially feasible solutions. In this stability, the future of nuclear disaster was understood as something represented more directly in the prediction map, not as a perceptual gestalt. Further, in the post-accident dominant stability, the pre-spectival focus has been not on the making-future-present process but on its output. Disruptive experiences in the acute phase of the Fukushima disaster, and political attention to the effectiveness of evacuation plans in the restarting process, led to a new self-determination network wherein the SPEEDI map is regarded as transportable (Crawford, 2005) and as a resource for self-determining evacuation. Thus, as the pre-spectival focus is narrowed and simplified (Callon, 1987), the complex and heterogeneous simulating radiological consequences become black-boxed or punctualized (Callon, 1986; Law, 1992) through the translation process, which results in a single-point actant: a prediction map. The delegated role of SPEEDI has been converted from supporting informed decision-making to prescribing evacuation behaviors. In turn, SPEEDI’s prediction map can be construed as a nonhuman actant that issues an order such as ‘go where I tell you’ to practitioners and residents.

OPEN IN VIEWER

Table 1.

Multistability of SPEEDI use.

	Designers	Pre-accident dominant stability	Post-accident dominant stability	NRA
Stability	Prediction as supporting advice	Prediction as a tool for evacuation drills	Prediction as self-protection	Prediction as a source of misunderstanding
Pre-spectival focus	Expert advice to decision-makers	Localization of prediction maps into socially feasible solutions	Prediction map itself	Difficulty of predicting accident progress preceding atmospheric dispersion
Form of life	Sensitive to the uncertainties of dispersion simulations as ‘environmental experts’	Focusing on societal uncertainties as local practitioners	Ordinary citizens’ sense that they cannot rely on the government at the critical moment	Putting emphasis on addressing squarely the uncertainties of reactor behavior as ‘reactor experts’
How to address uncertainties	Need for high degree of expertise	Manageable by keeping a margin in zoning	Easily understandable without scientific expertise	Almost impossible to predict the progress of a severe accident
How to decide evacuation	Informed judgment of the prediction and other inputs	Taking societal factors into account based on scientific simulations	Improvisation by residents referring directly to the prediction maps	Triggered mechanically by observable parameters
Actor-network	Social control network	Social control network	Self-determination network	Social control network

Against this trend, the NRA experts criticize what they see as reckless uses of predictive simulations, given their pre-spectival focus on the difficulty of predicting accident progress before the atmospheric dispersion. Their abolition of SPEEDI may be understood as a reaction of the social control network against the challenge of the new rival network. To uphold the agenda of social control, they cancelled the enrollment of SPEEDI, whose interpretive ambiguities were thought to be a source of misunderstanding and controversy. In this way, SPEEDI has been officially abandoned without re-opening (Lowe, 2001) it for public scrutiny. Rather than political secrecy, the SPEEDI controversy may come from the diverse pre-spectival focus on the making-future-present process reflecting the forms of life nourished through each group of actors’ professional and daily lives.

Additionally, the rivalry between two actor-networks can be considered a struggle over who uses the technology. When developed, the direct user of SPEEDI was assumed to be only limited environmental experts, while the residents were positioned at the very end of the supposed chain of calculators, advisors, and decision-makers. The same goes for the pre-accident dominant stability, while the role of advisors becomes invisible. In contrast, the emergence of a self-determination network stemmed from residents directly claiming use of the prediction map without screening by experts and decision-makers.

Note that, for both pre-accident and post-accident dominant stabilities, whether those ways of using prediction can truly operate has not been empirically verified. Among a hundred prediction maps calculated by SPEEDI during the acute phase of the accident, reluctantly disclosed by the Japanese government two months after the accident, some maps seemingly exhibited a similar pattern to actually observed dispersion of radionuclides. However, no one at that time could know which one was ‘correct’ (Juraku & Sugawara, 2021; Sugawara & Juraku, 2018). Hence, the post-accident dominant stability can be construed as an idealized construct in a retrospective discourse. The widespread use of SPEEDI despite the fact that the assumptions underlying the post-accident dominant stability remain untested is surprising. A deep-seated conflict around the image of the public and evacuation may be behind it.

Contested imaginaries of residents

Interestingly, Table 1 shows that, despite their contradictory attitudes towards using this technology, SPEEDI designers and the NRA are similar in what they require residents to do in emergencies. As described above, SPEEDI was originally developed to prevent the public’s mental turmoil and ensure social control. Similarly, NRA’s new framework strongly requires people to behave in orderly ways. Although emergency decisions have shifted from depending to not depending on prediction maps, the self-determination and voluntary actions of local people should still be avoided in a nuclear technocracy. In a post-accident self-determination network, however, the local public is constituted as autonomous individuals; thus, SPEEDI is expected to support their decision-making. In this network, evacuation is enacted as a manifestation of voluntariness to protect oneself from radionuclides’ predicted diffusion.

A focal point where the difference between these two actor-networks is observed is whether low-dose exposure can be tolerated after releasing radioactive materials. The current scheme stipulates that it is not until the measurement values of radioactive environmental monitoring are above a certain number that people in the UPZ can start evacuating. From the standpoint of emergency managers, even after the release of radionuclides, the level of radiation exposure is low, with an almost negligible health impact, which is greatly outweighed by the risk of societal disruption owing to so-called shadow evacuation. In contrast, for those who live in the UPZ, this new scheme cannot be logically and emotionally justified. It may be difficult for them to accept that they are forced to be exposed to radiation and not evacuate before release. They would prefer to use SPEEDI maps to evacuate proactively, rather than wait for instructions.

This inversion of the role of SPEEDI, from a supporting tool for ensuring social control to a primary means for enabling self-determination, may resonate with the emergence of citizen science after the Fukushima disaster. Several citizen-led radiation monitoring and mapping projects have been launched in response to the inadequate government transparency and communication regarding radiation levels (Abe, 2014; Brown et al., 2016). Likewise, SPEEDI might be understood as possibly contributing to citizen science as ‘an alternative locus of expertise less dependent on state and corporate actors’ (Stevens & Haines, 2020). This is an ironic consequence, given the original context of developing the SPEEDI.

Without reflexive perspectives that some scholars in natural hazards and disaster risk reduction hold, ‘participation’ in nuclear disaster management has been aimed, both before and after the Fukushima disaster, towards fixing the relationship between those who issue evacuation instructions and those who faithfully follow them. The public relations brochures for disaster risk reduction published by the Japanese government epitomize this difference. The brochure for natural hazards and disaster risk reduction including flood and landslide emphasizes the message of ‘protecting yourself’ and the need for proactive evacuation behaviors based on the residents’ own judgment (Public Relations Office of the Government of Japan, 2022). Meanwhile, the brochure for nuclear disaster management underlines the message of ‘there is no need for urgent evacuation’ and ‘it is important to behave calmly’ (Cabinet Office of the Government of Japan, 2020). Although the necessity of the ‘all-hazard’ approach is also advocated in the nuclear field (Nuclear Energy Agency, Organisation for Economic Co-operation and Development [OECD/NEA], 2018), it should be noted that the enacted images of the public and evacuation are fundamentally different between nuclear and natural hazards.

Discussion

The SPEEDI controversy derives not simply from issues of secrecy or communication but from more deep-rooted factors, namely the multiplicity of human–technology relations around predictive simulation, and the difference in the values prioritized under nuclear emergency. Hence, aiming to solve this controversy by improving the communication of predictions without addressing these deeper problems will elicit similar confusion in the next nuclear accident. It would not be wise for scholars of social sciences and humanities to escalate the conflict by taking a particular position without profound understanding. Instead, the practical contribution which STS scholars and postphenomenologists can make is to deeply investigate technology and society, to closely examine the background of problems, and to provide a basis for further public scrutiny and discussion, as this study attempts to do.

It is understandable that Japanese nuclear practitioners intended to optimize evacuation based on predictive simulations reflecting densely populated conditions of the country. However, the failure of institutional design for taking proper account of the nature of future that this simulation technology depicts may be one of the latent conditions for the ensuing SPEEDI controversy. If nuclear experts and practitioners at the time had constructed a nuclear emergency framework for adequately covering the uncertainties associated with predictions of the accident and radionuclide dispersion, the social consequences of the Fukushima nuclear disaster could have been different. ³ Paradoxically, this study suggests the difficulty of uncovering the deep-seated problems behind the use of predictive technology without probing into its philosophical and socio-political dimensions. Given this difficulty, it may not be realistic to allocate responsibility only to the technology developers and institutional designers. A large literature in STS has shown, sometimes critically, the need for ‘real-time technology assessment’, ‘upstream engagement’, and similar (e.g., Guston & Sarewitz, 2002; Stilgoe et al., 2013; Wynne, 2002). More profound STS and philosophical interventions will be needed in the development and institutionalization stages for technologies like SPEEDI that are rarely used but may affect many people. While further research is needed for putting it into practice, the following four points regarding the theoretical and methodological challenges for postphenomenology-ANT studies may provide some basis for its realization.

The first one is concerning the issue of materiality around predictive technology. The concept of pre-spectival focus suggests that the ways humans give attention to the making-future-present process and what kind of materiality of future-world-and-technology amalgamation is highlighted may co-shape the multistability of SPEEDI use. As these stable usages are linked with different actor-networks in which the constituted public and enacted evacuation under nuclear emergencies vary considerably, different, even opposing, roles are delegated to SPEEDI in each network. Such a postphenomenological–ANT account of the SPEEDI controversy provides an opportunity for rethinking the socio-centric account that reduces it to a simple problem of political and bureaucratic secrecy and for seeking the right balance between the social and material. Rosenberger (2018) has already asserted that articulating the nature of technological materiality is vital for the further amalgamation of multistability studies and ANT. Although that is compelling, we need to be careful when discussing the materiality of predictive technology. In the case of the physically present artefacts, materiality may be discussed straightforwardly. However, in predictive technology, which is already a constituent of future-world-and-technology amalgamation, which parts are the focus and how materiality is relevant depends on the interplay between human perception and the world-technology entanglement. The diversity of such interplay is a crucial source of multistability in using predictive technology. For further postphenomenological examination of predictive technology, it will be a crucial challenge to elucidate how its technological materiality is molded in conjunction with theoretical discussion around the material and materiality (Law, 2010).

The second is the distance between the technology usage assumed by the designers and the dominant use of the technology. Conventionally, multistability studies have provided possible stabilities, other than the predominant use, to critically examine things that are otherwise invisible or suppressed unjustifiably, such as the anti-homeless design of public benches (Rosenberger, 2014). One of the definitions of dominant stability as ‘the one for which it was designed and manufactured, and the usage for which it tends to be taken up’ (Rosenberger, 2020a) shows that these strands of studies have nearly equated it with the designers’ original intention. It has often been taken for granted that technology designers have privileged positions in the entire technology usage process. This may lead to critical perspectives against the ‘systems of marginalization’ where ‘those in marginalized positions have a special view of these mechanisms, and those in dominant positions may fail to recognize them owing to their privilege’ (Rosenberger, 2020a). However, the SPEEDI case may cast doubt on this assumption. The dominant stabilities of SPEEDI use, regardless of whether they were pre- or post-Fukushima, do not fully reflect what its designers had assumed. Instead, its intended usage has been structurally ignored in the main discourse on nuclear emergency management (Juraku & Sugawara, 2021). This signifies that technology designers do not always retain the privileges by which they can dominate its usage. Furthermore, it is not always true that the legal authority’s standardized usage accords with the dominant public imagination of the technological use. It is evident from the phenomenon that the public voice for the proactive use of SPEEDI has repeatedly been raised, despite its exclusion from the official emergency management framework established by the regulatory authority. ⁴ This may imply the need to examine the extent to which the dominant stability reflects the intention and expertise of designers and envision possible alternatives. Although postphenomenology, relinquishing essentialism, does not take the stance that the usage for which it was designed is the only ‘correct’ one, at least the distance between design and dominant use(s) should be squarely scrutinized.

A third point is the need to expand what Rosenberger calls ‘variational cross-examination’. It involves critically contrasting the various stabilities of a multistable technology with the purpose of ‘exploring how a particular stability has come to dominate’ (Rosenberger, 2014, p. 369). When examining predictive technologies whose usages are not experientially established, it may be crucial to take full advantage of multiple perspectives and insights acquired through the multistability analysis to test whether each stability can operate fairly and reasonably. Particularly for SPEEDI, it is difficult to establish a learning loop with lived experience feedback because nuclear accidents are infrequent.

As discussed above, the SPEEDI controversy is thought to partially originate in the distance between the designed and the dominant uses. SPEEDI was institutionalized in a form detached from the designers’ intention, without ascertaining whether it could be genuinely used in such a way. Considering the possibility of tremendous variance in the direction of atmospheric dispersion depending on the timing of release, it may be questionable whether the pre-Fukushima dominant usage adequately addressed the uncertainties of simulations. Similarly, while valuing self-determination is desirable in the post-accident dominant stability, it is worth investigating whether using SPEEDI as a direct resource for self-determining evacuation would bring serious disadvantages, such as increasing radiation exposure. Such investigation should constitute the basis for public discussion on how to strike the right balance between minimizing radiation exposure and ensuring self-determination under nuclear emergencies. In more general terms, ‘variational cross-examination’ may well complement the lack of first-hand experience when envisioning the usage of predictive technologies, if adequately expanded, by examining the extent to which each stable use is technically and socially feasible, conditions under which each stable usage can effectively operate, and discussing which stable usage can be more ‘adequate’ from multiple perspectives.

And last, the SPEEDI case may suggest a new perspective in the philosophical and practical account of ‘learning’ the technology. Conventional postphenomenology and STS have seemingly construed learning as becoming ‘transparent in use’, such that humans can use the technology without consciously seeking ways to handle it. A hearing aid can be a (quasi-) transparent technology for those who usually wear it (Ihde, 2009), while many people do not reach the same level of transparency and embodiment (Rosenberger & Verbeek, 2015). After updating or refurbishing a technology, efforts are made to make it perceptually transparent again (Forss, 2012). Such views on ‘learning’ may also be shared with Bourdieu’s habitus, which highlights the unconsciousness when a set of dispositions, schemas, and ways of knowing are deeply internalized (Bourdieu, 1990; Swartz, 2002). The SPEEDI case may suggest the opposite view. The more (scientific or technological) expertise the actors have, the broader focus they seem to consciously have on the uncertainties and conditionalities of the making-future-present process. On the contrary, for the actors such as local practitioners and residents who have less expertise in predictive simulation, the complex prediction process becomes transparent. In this case, developing expertise may function as resistance to simplification, making it transparent. Thus, elaborating the analytical concept of pre-spectival focus will provide broader and richer accounts of what ‘expertise’ and ‘learning’ mean and the mechanism of making prediction outcomes transportable out of their contexts wherein they are generated. For example, it may be useful to apply this analytical concept in the study of human–technology relations and what ‘learning’ means in case of predictive technologies similar to SPEEDI, including ARAC in the US, real-time on-line decision support system (RODOS) in Europe (Ehrhardt et al., 1993), and more simplified simulations such as Outil de sensibilisation aux problématiques post-accidentelles à destination des acteurs locaux (OPAL) in France (Jolivet, 2012). Such comparative analysis will reveal the similarities and differences of the pre-spectival focus among different societies, which will provide further insights on how differences in technical details of predictive simulations and in the socio-political contexts in which they are embedded may relate to human–technology relations.

Conclusion

Although SPEEDI was expected to provide real-time predictions of radiological consequences, it could not meet this expectation during the first days of the Fukushima nuclear disaster. Japan has become involved in a socio-technical controversy over its usage in the lessons-learning process. A socio-centric account has become publicly dominant: SPEEDI, despite providing a series of prediction maps, had been underutilized owing to political and bureaucratic secrecy. In contrast, regulatory experts banned the use of SPEEDI because they thought that it is almost technically impossible to predict the behavior of a damaged reactor in a severe accident. This article analyses this controversy as an issue of multiple human–technology relations around predictive technology, keeping distance from both socio- and techno-centric explanations, and explores it through the concept of multistability in postphenomenology.

For a postphenomenological study of predictive technologies, the author introduces a new concept of pre-spectival focus, which refers to how human attention is directed to the making-future-present process and which features or aspects of its process are foregrounded or backgrounded. It explores how humans may perceive and experience the complex process of rendering future processes composed of chains of technologies and human interventions. Working with this conceptual lens and ANT, the multistable usages of SPEEDI can be summarized as the following four: prediction as supporting advice, prediction as a tool for evacuation drills, prediction as self-protection, and prediction as a source of misunderstanding.

Initially, the pre-spectival focus of SPEEDI designers was on the need for prudent interpretation of predictive maps and providing expert advice to decision-makers, due to the limitations of simulation. They construed the future of nuclear disaster as a perceptual gestalt that emerged by integrating various information and hermeneutic interpretation. As SPEEDI became institutionalized in the nuclear emergency drills, the pre-spectival focus of practitioners had shifted from expert advice to the tuning process of prediction maps reflecting local conditions. While the uncertainties associated with predictive simulation were relegated into the background, SPEEDI maps were regarded as scientific evidence directly prescribing evacuation areas. Third, in the post-Fukushima publicly dominant discourse, SPEEDI maps were presumed to be available immediately for the Fukushima residents to refer to directly for their self-protection. In this stability, the pre-spectival focus was put on the prediction maps themselves. As those maps are understood as representing neutrally what will happen in the near future, expert interpretation is deemed akin to arbitrary screening or censorship. In contrast, by putting their pre-spectival focus on the difficulty of predicting accident progress, the NRA experts have a sense of danger against the reckless use of SPEEDI. Eventually, they inhibited the use of this technology and introduced a new framework that did not depend on prediction.

Despite seeming sharply opposed, the pre-accident dominant stability and the NRA’s new emergency management framework have a commonality in being linked with social control network: the residents are constituted as passive selves who obediently follow the governmental instructions. Rather than the superficially regarded contestation over the use or non-use of SPEEDI, a distinct rivalry can be observed between the social control network and the self-determination network, in which residents are constituted as autonomous subjects who can actively decide protective actions. In these two actor-networks, the delegated roles of SPEEDI are almost opposite, a tool for ensuring paternalistic social control or a primary means for enabling self-determination. This rivalry around the enacted images of residents and evacuation may make the SPEEDI controversy smoldering and complicated. A postphenomenology–ANT examination can help create awareness of values otherwise closed off in the dominant stability of technological usage; in this case, the issue of self-determination under nuclear emergency in the shadow of a taken-for-granted purpose of minimizing radiation exposure.

Using SPEEDI in both pre- and post-Fukushima dominant stabilities has not been tested by first-hand experience of nuclear disaster. Postphenomenological ‘variational cross-examination’ may well complement such a lack of bodily experience and feedback loop, if properly expanded, by examining how each stable use is technically and socially feasible and discussing which stable usage can be more ‘adequate’ from multiple perspectives.

Postphenomenological studies have traditionally treated multistable usages as equal (Ihde, 1990), such that ‘of which we cannot say that one is more appropriate than the other’ (de Boer, 2021). However, instead of refraining its critical analysis from stepping into the intercomparison of multistable usages, it will be beneficial to reify philosophical interventions into constructing better use of predictive technology.