Accident Report Interpretation

Abstract

The language and approach we use to describe the past can have a strong influence on the audience’s interpretation of our story. In our experiment, we explore, using 3 different conditions, how the framing, language and style of an accident report can affect the audience’s proposed solutions to manage the problems found. We find that the approach used to create an accident report can have a powerful influence on the audience’s decision making. Whether we are describing an accident in a linear manner, using a systems approach, or we are accepting of multiple stories which are not linear or coherent, the methods we use to capture and communicate the story have a profound impact on the actions decided upon by the reader.

1. Introduction

How stories are framed can greatly influence the readers’ interpretation of the event and the actions taken as a result. The analysis of accidents and the accident report style used to share the story provides a prime example of this.

The language we use to tell a story can have a profound effect on how the listener or reader interprets the story. The power of language and metaphors is showcased in the work of Thibodeau and Boroditsky [1]. In their study, Thibodeau and Boroditsky presented participants with different descriptions of the same fictional city suffering from a surge in crime. Their study sought to understand how metaphors shape understanding and reasoning, and whether the use of different metaphors to describe crime would lead to the proposal of different solutions. Their study’s results showed that when the word “beast” was used to describe crime, the proposed solutions were much more likely to be enforcement-based, while when the word “virus” was used to describe crime, the proposed solutions became much more social-reform based.

The influence of the English language on our thoughts and actions, as opposed to other languages, should not be understated either. In the description of accidents, the English language consistently uses the agent of causality to describe the event e.g., “John dropped the vase”. Conversely, when describing accidents in Spanish or Japanese, the agent of causality is dropped, “The vase broke itself” [2]. This showcases how those of us who speak the English language are more likely to naturally focus upon human error because of our instinctive way of describing the world we see. When this is refined to how we normally hear about accidents, which is usually in a headline format to keep it sharp and simple, the accident becomes more about the individual than about the system issues which have led to it occurring due to how we tell the story. This can often unintentionally assist the creation of a blame culture and can have a serious impact on an organisation’s ability to learn.

Although the world of academia has generally accepted that most industrial accidents occur due to complex system issues [3] and that human error is but a symptom of a larger problem within the organisational system [4], many organisations, and society, remain deeply focussed on the front-line operator after an accident has occurred. Many accident reports still cite human error as a primary cause [5]. Reports for notable accidents such as the Bhopal chemical plant [6], the Herald of Free Enterprise [7] and the Kegworth air crash [8] have all highlighted human error as a key factor in the accident’s occurrence. All 3 of these accidents are used within safety training curriculum as examples of how things go wrong to this very day. This leads organisations to believe that it is the operator who is the main contributor to accidents and it is the operator who is the problem within their system which requires rectification. Today, almost every accident is followed by questions focussed on ‘‘whose fault?’’ [9] It seems as if every accident must be charged to somebody’s account [10]. This often results in someone being blamed and punished for an accident which was essentially created by complex system problems which arose from various relationships and interactions within the system [11].

The use of punitive or blame methods to manage mistakes or errors can be not only harmful to the worker [12] but also to an organisation’s learning ability [13]. Workforces are much less likely to report mistakes for fear of retribution [14,15], creating an organisational culture where a chief executive officer (CEO) only learns of the problems within their organisation after a person is seriously injured. Using punishment to deal with error is likely to create a culture where the workforce resent the very system that was supposed to enable them to work safely because of it using them as the sacrificial lamb to appease society when something goes wrong. The greatest impediment to learning from mistakes is the use of punishment [16] as you cannot have a system which punishes those who make mistakes while also trying to maintain a learning culture [17] which needs people to be open and honest when an error is made.

To counter these harms within our systems, the use of restorative mechanisms has been promoted as a more beneficial alternative to manage failure in the workplace and to achieve justice. Rather than continuing to use solely retributive mechanisms that will create long-lasting workplace wounds, the use of restorative mechanisms is proposed as a means of “addressing the harms that have been done and addressing the causes of these harms” [18]. Using this approach, the line of questioning following an accident would no longer be who is responsible, but what [19], and how can we fix the issue; not the person. The restorative principles, particular those put forward by Zehr and Gohar, are aligned with the work of Doctor Albert Wu and his Second Victim Theory [12]. Wu’s theory highlights the impact an accident has upon those who feel personally responsible for the accident’s occurrence and the need for organisations to help these individuals to recover from their mistake, rather than vilify and ostracise them.

The accident model and writing style chosen by the accident facilitator play a critical role in the level of learning achieved by an organisation following an accident. There are many accident models available today, with several having contrasting styles. Some follow the root cause analysis approach, where the use of reductionism can often lead to a form of human error as a conclusion. Other models are more systems-orientated, placing the focus more on system relationships and emergent risk. For the accident facilitator, “what you look for is what you find” [20]. If you are looking for someone to blame, they are easily found, particularly in hindsight. The style of model used may dictate the information provided to the reader. For an organisation deciding upon the actions to prevent a similar reoccurrence, “what you see is all there is” [21]. The reader’s understanding of the event and their opinion of the actors involved is heavily reliant upon the information provided within the accident report and the framing and language used by the accident facilitator.

In this paper, we empirically analyse whether using different metaphors and writing styles to tell the accident story leads the reader to focus on different aspects of the accident report and to propose different solutions. Will the solutions vary in line with the writing style? Will the metaphors used encourage people to reason about the event in a consistent manner with the underlying perception of these metaphors? For example, if we use language which is consistent with how law enforcement describes a crime, will people treat the event in a similar manner to how crime is dealt with in society? Will the focus of the accident analysis facilitator and their chosen facts influence the reader? Or if the event is told in a way which is aimed towards a learning outcome, will the solutions proposed be consistent with the aims of the storyteller?

2. Materials and Methods (A Study of an Accident Report Interpretation (Griffith University Ethics Approval Ref. No.: 2017/839))

2.1. Participants

For this experiment, we used a sample population of 93 people, all of whom work for construction companies. The study was issued to the operational staff members of three different companies and the studies were issued in alphabetical order using each organisation’s distribution list. Our sample population included various sub-populations such as engineers, construction managers, human capital personnel and safety advisors. This allowed us to sample a broad cross section of the production support team. This was necessary as people’s conceptions of workplace accidents may differ depending on their past experiences, their role and their exposure to the accident process.

The data was collected using both paper and online anonymous surveys and there was no incentive to participate. To preserve the confidentiality of the participants, no individually identifiable information, including the participant’s organisation role, was collected.

2.2. Materials

The experiment involved 3 conditions, titled “Report Variants”. Each participant was randomly allocated 1 of 3 accident reports, such that each report was reviewed by one-third (31) of the participants. In each of the 3 reports, participants were presented with a description of an event which had the potential to cause life-altering harm to human life. Each report described the same event. The reports differed in their language, structure and style. Report Variant 1 used a reductionist approach to create a linear story; Report Variant 2 focussed on issues existing within the organisation’s system; and Report Variant 3 focussed on the provision of multiple accounts.

The facts included in the reports used were all taken from a real-life accident. Due to the different report styles used, not all facts are captured in each report. This was natural rather than forced due to the accident report style dictating the information captured and the way it was presented.

Each report information sheet requested the participants to “review an accident report and provide 3 recommendations based on the information provided”. The heading above the 3 blank boxes provided asked the participant to “identify preventative/corrective actions”.

Participants could access the report and submit their response in either paper or electronic format.

2.3. Conditions

2.3.1. Report Variant 1

The search for human error following an accident has long been a goal of many accident analysers. One of the factors which greatly increased the focus on human error following an accident was the event that occurred at Three Mile Island and the use of the label “human error” in the subsequent report [22]. The spotlight on human error within the world of accident analysis intensified in recent decades because of human error-focussed research, such as that of Stanton and Glendon’s work on individual attitudes to risk taking [23]. This led those within the field of accident analysis to place a disproportionate amount of time focussing on human issues and viewing them as the main cause of accidents. Another influencing factor in industry choosing this trajectory for the analysis of accidents was the work of Heinrich, who declared that 88% of all accidents are the result of human error [24]. A key issue with focussing upon the symptoms of human error is the underlying causes which helped create the error are ignored [25], as well as the weaknesses in the organisational and regulatory system which created the environment for the error [26]. These less direct forms of human failure help to establish the working practices that result in operator error [27].

The report used for Report Variant 1 was a real accident report written after an accident occurred on a construction project. It was modified only by removing identifying details. The writing style used by the original report’s author is human-error focussed, with minimal focus on the underlying causes of the errors [Appendix B]. In creating the report, the accident analysis facilitator reviewed the interviews of those involved and selected the information they believed to be relevant to the report.

2.3.2. Report Variant 2

There are many safety researchers who encourage the operational world to move away entirely from a focus on human error and towards a complete system focus. Accidents today are much too complex to understand through traditional accident analysis, which aspired to find individual broken components such as human error [11]. System behaviour cannot be deduced from component behaviour [28] as “simple entities, because of their interaction, cross-adaption and cumulative change, can produce far more complex behaviours as a collective” [29].

There are many methods used in the world today to analyse accidents using a systems approach. The Functional Resonance Analysis Method (FRAM) [30], Accimaps [3], Systems Theoretic Accident Model and Processes (STAMP) [11] and Strengths, Weaknesses, Opportunities and Threats (SWOT) Analysis are just some of the processes used to better understand complex accidents.

The report used for Report Variant 2 was based on the same accident used for Report Variant 1. This report was written by the first author for this experiment and contains information which was not included in Report Variant 1 because of the writing style chosen. This additional information is also factual. Participants were provided with a one paragraph summary of the accident followed by a SWOT Analysis of the 5 elements deemed necessary to achieve success [Appendix C].

These elements were:

• Daily activity briefing;
• Personnel;
• Tools and equipment;
• Work environment;
• Task execution.

2.3.3. Report Variant 3

Traditionally, accident reports consisted of a singular narrative describing the event. The role of the accident analysis facilitator was to objectively collate the facts related to the event and to provide a story for others to better understand what happened and how. The accident analysis facilitator would choose what they believed to be the important aspects from each statement of those involved in the event and use this information to create a linear story. A key failure in this approach is that it is impossible for the facilitator to avoid hindsight bias using this method, given that they are aware of the outcome of the event. Hindsight allows observers to change past indeterminacy and complexity into order, structure, and oversimplified causality [31]. The facilitator regularly omits background information within each actor’s story from the final report to help create the story they believe to be true. This story is often littered with counterfactuals highlighting what the operators could or should have done. However, saying what people could have done to prevent a specific outcome does not explain why they did what they did [32].

The report used for Report Variant 3 again communicated the same accident as Report Variant 1 and 2. This report, written by the first author for this experiment, used the “multiple stories” approach [Appendix D]. This report, like Report Variant 2, contains information which was not included in Report Variant 1 because of the writing style chosen. This additional information is also factual. Each actor’s account was recounted verbatim in the final report to ensure everyone’s voice was heard, whether conflicting or not, and this became the accident report. The issues and constraints faced by those involved and captured in their story were listed as contributing factors to the accident.

3. Results

Overall, there were 298 corrective actions proposed by the participants across the 3 conditions; 99, 100 and 99.

3.1. Coding

Proposed corrective actions were coded into 2 categories: (1) Human/blame-focussed; and (2) system-focussed. Each category was further broken down into sub-categories. Human/blame-focussed was divided into the following elements: (1) punish the people involved; (2) non-punitive action focussed on the people involved (e.g., training, reinforcement of correct behaviour). System-focussed was divided into the following groups: (1) one-off actions such as communicating about the accident or reviewing the risk register considering the accident; (2) changes to the physical workplace; (3) reinforcement or change to the work practices specifically involved in the accident; (4) reinforcement or change to practices not directly involved in the accident; and (5) changes to documents. A separate category was provided for any counterfactual statement found. This category was then separately coded using the 2 main categories.

Each participant’s response was weighted equally—as a single point towards the analysis.

Participants’ corrective action suggestions were coded blindly by two coders. Cronbach’s Alpha—a measure of reliability—was 0.957759, indicating good agreement between the coders. The coding made by the first author was used for the remainder of the analysis.

3.2. Findings

Overall, participants were more likely to propose system-focussed corrective actions (255 actions, 86.44%) than human error-focussed (40 actions, 13.56%). However, the actions participants proposed to manage the issues which led to the accident varied greatly depending on the style of the report, the framing of the story and the language used to describe the event. Participants provided with the traditional approach report were more likely to suggest actions with a human error/blame-focus (27 actions, 27.55%) than those provided with the systems approach report (8 actions, 8%) or the multiple stories approach (5 actions, 5.15%) reports, χ²  =  25.39, p < 0.001, see Figure 1.

As seen in

Table 1. The table below displays the distribution of accident classifications across the 3 report variants.

Classification	Report Variant 1	Report Variant 2	Report Variant 3
Non-punitive action focussed on the people involved (e.g., training, reinforcement of correct behaviour).	8	2	1
Punish the people involved.	4	0	0
Counterfactual statement—Human Focus.	15	6	4
Counterfactual statement—System Focus.	1	12	12
Changes to documents.	1	0	3
One-off actions such as communicating about the incident, or reviewing the risk register considering the accident.	6	10	9
Changes to the physical workplace.	7	16	6
Reinforcement or change to practices not directly involved in the accident (e.g., toolbox talks, site coordination, competency management).	32	28	50
Reinforcement or change to the work practices specifically involved in accident (protecting against dropped tools).	25	26	15

, readers of Report Variant 1 were the only participants from all 3 reports who requested punishment for the individuals involved (4 actions, 4.04%).

Appendix A contains examples of the recommended actions in each classification.

4. Discussion

4.1. Interpretation of Results

The results show that participants presented with Report Variant 1 were more likely to recommend actions that are directed at front-line operators than participants presented with Report Variants 2 and 3. Only readers of Report Variant 1 chose punishment as an action. Although the number of such cases is small (4), it is significant that none of the readers of Report Variants 2 or 3 chose this type of action.

In terms of reaching a conclusion from our results, there are 3 possibilities:

• The results are random and do not mean anything.
• Different information within each report influenced the participants.
• Different styles used to present the story influenced the actions decided upon.

The reports were issued to a wide range of individuals each of whom operate within a specific discipline in the construction organisation. Each participant will have had some form of exposure to a workplace accident prior to participation in this study as all of the disciplines were chosen based on their exposure and influence upon the accident process within the organisation. The results, rather than having been influenced by the style or language of the reports, may have been instead dependent on the reader’s personal opinion of accident causation and the type of action required following any event. The reader may have used their previous knowledge and experience to assess the information provided and chose their actions accordingly. The actions which were human and punishment-based may have been chosen by the participants who believed their selected actions would help the organisation and the people involved. It may have been by chance that many of the participants who believe a focus on the human is beneficial received Report Variant 1 rather than Report Variant 2 or 3.

When carrying out these type of experiments, a participant’s past experience and their personal beliefs will have some form of influence upon their decisions and actions. For example, a reader may see similarities between the accident report information and a previous accident they were involved in and decide upon similar actions because they believe these actions were successful previously. The participant’s beliefs may also play a role. A person’s beliefs are shaped by society, religion and previous experience. It is a strongly held belief among many that a mistake which did or could have caused serious harm requires retributive action to ensure the human involved pays what they owe.

Whilst past experience and personal beliefs may indeed affect decision making, this is insufficient to explain the results. Out of the 3 reports, only Report Variant 1 is human error-focussed. This report was also the only report to be provided with actions that recommended the use of punishment. Although the number of punishment recommendations was small, this is still statistically significant. Were the results truly random and based only on past experience and personal beliefs, there would be a high likelihood that the human-focussed actions would be more evenly distributed or higher in Report Variants 2 or 3 as opposed to Report Variant 1.

Report Variants 2 and 3 both have background information provided within them that does not appear within Report Variant 1. The provision of these additional opinions and facts may have helped sway their readers’ focus from the individuals involved towards system fragilities. Conversely, had this information been provided within Report Variant 1, the number of actions focussed on the front-line operatives may have dramatically decreased, as the reader would have been given a more comprehensive picture of the accident and the reasons why people did what they did.

Report Variants 2 and 3 did have additional background information provided in several areas where Report Variant 1 did not. A key reason for this is that the accident report style chosen dictates the content of the report. The style of Report Variant 1 is heavily reliant on the accident facilitator’s perception and interpretation of “important” facts. Given that as humans we are naturally susceptible to hindsight and confirmation bias when analysing past events, the style of Report Variant 1 makes it more vulnerable to omitted information than Report Variants 2 or 3, as any conflicting information creates a choice for the facilitator to ensure a linear story is achieved. As the key goal of Report Variant 1 is to put forward the proven facts as the facilitator sees them, the background information captured within Report Variants 2 and 3 is often written off as opinion and hearsay. Reports based on provable facts can help an organisation understand what happened, but they are far less effective in helping us understand why. Human error is easy to “prove” in hindsight and with the help of post-accident procedure reviews, whereas problems within a system can be much more difficult to label as “broken” because often system issues exist among a variety of systems that are individually working as designed. It is only their interactions with other systems that are creating the unmanaged risk.

The accident is presented in 3 very different styles. Report Variant 1 prioritises facts and minimises any background information deemed irrelevant by the author. Report Variant 2 delves into different facets of the system and the elements that were successful against those which were creating unmanaged risk. Report Variant 3 seeks to allow the various actors involved in the accident to be heard to ensure the reader has a full understanding of what happened and the constraints faced by each of the actors. The report style chosen has influenced the focus of attention. Report Variant 1, due to its focus upon what went wrong, places a strong focus upon the front-line operatives which may have led to the some of the report’s readers to decide that the accident was the result of careless workers. With Report Variant 2, the focus migrates from the human to the system. The description of the issues found are more focussed on the “what” than the “who”. This in turn may have led the reader to provide actions based around what needs rectification, rather than who. Like Report Variant 2, Report Variant 3 ensures the human is not the focus of attention. Each actor’s story, and the constraints they faced leading up to the accident, is heard in full to ensure these people are understood and the reader gains a deep understanding of the world these people operated within at the time of the event.

Although these reports are related to the same accident, each of them has slightly different objectives. Report Variant 1 uses a reductionist-style approach to find the broken parts that caused the accident and to fix them. Each fact is assessed against the safety management system to determine whether it is “broken” or not. This creates a binary view of the accident and the actions of those involved. This in turn can lead to a “guilty” or “not guilty” description of the actions of those involved. Report Variant 2 focuses solely upon the system and the various relationships and issues that exist within it. A human mistake is viewed as something that has been caused by a bigger issue which exists within the system. In a similar manner to Report Variant 1, this report can transform what began as an accident which was highly personal to all of those involved into an objective, lifeless narrative which fails to capture the mindset of the accident’s actors. Report Variant 3 is focussed upon allowing the actors to be heard so that any damage suffered by them can be healed through the disclosure of their story. The front-line operatives are viewed as victims who require support. These differences in objectives can have a substantial impact on the type of actions determined by the reader. One of the consequences of the different styles of reporting is both the author and the reader being drawn to different information. Report Variant 1 is over-reliant on the author needing to view information sourced from those involved in the accident as important for it to be included within the final report, whereas Report Variant 3 is much less reliant on the author’s own views and is accepting of contradictory facts without resolving to a single conclusion. This highlights how careful organisations need to be when selecting their accident report style and whether the style aligns with their objectives. The results of our study suggest that information viewed as unimportant, not relevant or otherwise by the report’s author, can have serious implications for those involved in accidents and the ability for the organisation to learn from its mistakes.

A limitation of this study is the fact that it was conducted in abstract, as the participants were engaged in an experiment rather than a real accident review. This creates the potential for participants to respond differently to how they would respond to an accident report within their workplace because of the removal of any socio-political aspects which may exist in their organization.

4.2. Implications for Practice

The findings in this study suggest that the accident report style and model selected by an organisation or the accident analysis facilitator to analyse the accident can have a significant impact on the outcomes of the accident analysis. This has important implications for any organisation seeking to achieve a “just” culture. The accident process is a core element in establishing and maintaining a learning culture within an organisation. Learning from accidents, particularly those with serious consequences, often influences an organisation’s future safety strategy. The treatment of those involved in accidents can create or destroy the foundation of trust needed to support a positive reporting culture and the sharing of critical information within an organisation. Facilitators of accident analyses must accept that the accident report style chosen has serious implications for those involved and the future ability of the organisation to learn. Our results suggest that the pursuit of a linear report based only on facts determined as important by the author may increase the potential for recommended actions to be blame-focussed and impede the organisation from dealing with more serious issues which have been deemed irrelevant by the author. The omission of background information provided by the actors involved in the event from the final report can have severe repercussions for these actors and for future reporting within the organisation.

This study raises legitimate concerns that researchers need to do more work in the area of accident analysis to provide better advice to safety practitioners in future. As with accident reports themselves, readers should take care in under or over interpreting the results from this study. The results are strongly suggestive that accident report style influences accident analysis outcomes—but alternate interpretations can be drawn from the results. Further work is needed to establish conclusive relationships between the facts, methods, styles and outcomes of accident reporting.