The antecedents and development of unsafety

This study explored the conceptualisation and language of ‘unsafety’ based on a large dataset of nearly 4000 Safety Observation reports (SOR) from a large infrastructure construction project.

E.g. How do people identify and define safety, unsafety, unsafe acts and conditions based on content analysis of the reports.

While we often talk about ‘safety’, our practices conversely often focus on the absence of safety (via accidents and the like); hence, unsafety.

Providing background:

·???????? The historical development of safety in construction “has in part dictated its lexicon”. E.g. the early focus was on identifying physical risks and machine guarding, then eventually moving to people and unsafe acts; then moving through to the current day via waves of sociotechnical lenses, behaviour, climate/culture and more

·???????? “A consequence of this ‘language of safety’ is the way it has shaped and even directed safety management thinking and practice”

·???????? An example of how this language of safety directs practice is via accident and near-miss reporting, and its influence in the construction of unsafe acts or unsafe conditions “as a binary either/or situation”

·???????? Despite this apparent simplicity of the lexicon, “accidents are often highly complex in reality, and to use such a simplistic dichotomy in reporting and subsequent investigation is likely to limit the learning potential of an incident”

·???????? It’s said that while management trends may tend to focus on the ‘acts’, organisational factors “are often critical”. Moreover, “unsafe acts and unsafe conditions often remain segregated in practice, reinforced by the lexicon itself”

·???????? Alongside the idea of unsafe conditions is unsafe acts, e.g. behaviour of people. Quoting Hollnagel they observe “the idea that human error could be used to explain the occurrence of adverse events was eagerly adopted”

·???????? Developing technology and systems to be fail-safe so “unsafe acts cannot lead to an accident, rather than educate workers through training programmes” is a warranted approach, “this is highly problematic within the construction industry”

·???????? For instance, they observe a challenge that “the continuing development of technology within the workplace has led to increasing complexity and coupling between tasks and activities, therefore interactions cannot necessarily be fully planned, understood or anticipated”

·???????? Also, while prevention of single point failure mechanisms is “often built in to processes and equipment”, the long subcontracting chains and complex and coupled work means that “occurrences with two or more cumulative failures … are harder to predict and therefore harder to prevent”

·???????? Ideas of organisational failure and systemic safety are said to bring together unsafe acts and conditions. Unsafe acts may represent a “symptom of deeper latent problems within projects or organisations, the management system creating situations, or rather unsafe conditions that can encourage or even force human errors within certain contexts”; thereby effectively encouraging undesired performance variability, e.g. competitive tendering, production pressures and bonus schemes

·???????? Moreover in construction where constant workplace and project change is in effect, including variable resources, poor working conditions, tough environments and complex co-ordination of different trades, “performance variability can be argued to be a necessity, therefore to isolate and label unsafe acts within such (potentially unsafe) conditions seems incongruous”

·???????? But these considerations have not “stopped continued focus on unsafe acts, embedded as they are in the historical language of safety”

·???????? Further, accidents can be seen as evidence of error or failure, described in terms of unsafe acts or conditions, where investigations seek to identify human fault. This search “has perpetuated ‘human error’ as a prominent causal factor in accidents … as the cause becomes easily identifiable as [a type of violation]”

·???????? Despite this, “the quest for root causes has been challenged on a variety of levels, not least the potential for over-simplification”. Trevor Kletz observed that “root cause has an air of finality about it, not always helpful, given that the cause of many construction injuries is actually gravity”

·???????? Motives for cataloguing safety situations via safety observation reports, incident reporting etc. stems from a desire to measure the safety status of a project; allowing the production of statistics that “are often proudly proclaimed at the entrance to projects, that announce the number of days or hours worked since the last accident and allow contractors to measure ?themselves against industry metrics”

Results

Of the ~4k SORs: 2k were categorised as unsafe conditions, 697 as unsafe acts and >1k as a good practice.

An immediate observation is that >75% of reports were labelled as unsafe conditions, seen as “surprising”, and highlighting that the majority of unsafe observations are linked to situations/conditions not influenced by human actions.

From the below chart it’s evident that unsafe conditions exceeds acts in all categorise except for ‘behaviour’. They note that the inclusion of ‘behaviour’ as a category is “in itself is interesting – it is neither a work type (such as excavations or lifting) nor an organisational function (such as permits, PPE or welfare)”.

The authors flag behaviour as an ‘unusual tag for a category’ in their view, since all of the other categories in the SOR system were facets of the project, e.g. physical issues or processes.

Further, they found 48 observation reports that were recorded as unsafe conditions “but many of these did not fall under a definition of situations that were unsafe through non-immediate human means”; highlighting inconsistency in the definitions.

Further, the authors re-evaluated those 48 reports – resulting in 90% being changed by the authors. This illustrates “the complexities involved in deciding at what point an act, or number of acts, eventually emerges into a condition; these decisions are inherently subjective”.

Moreover, while these findings “arguably supports more complex, non-linear and emergent approaches to analysing safety incidents”, of potentially more utility would be “clearly defined and shared understanding of the ‘line’ between acts and consequentially emergent conditions”.

Finger pointing was also observed through the SOR data, e.g. “Safety rep parking vehicle in live traffic route to speak to his supervisor”. Examples highlighted individuals with some level of authority (safety people, security guards and more).

While many unsafe act observations just reported the behaviour of an unidentified person, 37% directly identified the person by name or company, registration of vehicle or more. This data suggests “highly complex social and organisational issues at play that have seeped into the safety observation process, in part those who create and enforce the policies are readily punished by others for their violation”.

Both behavioural acts and conditions were easily assigned to either policy, procedural or equipment categorisations (see image below). They believe that a more useful categorisation could be made at a more detailed level within the data rather than relying on the traditional act/condition dichotomy.

Themes of ‘deliberate’ and ‘inadvertent’ also emerged as key categorisations. Deliberate was supported by lexicon of shortcuts and deliberate violations.

Further inconsistencies with categorisations emerged when evaluating hot work safety observations, particularly in the behavioural category. They observe that at least 50% were “incorrectly categorised as unsafe condition when could be more appropriately labelled unsafe act” and the reports that directly identified individuals or companies “could be considered ‘blame reports”.

These examples are said to highlight the ‘distance’ between the unsafe condition as it’s recorded and the unsafe act that led to the condition.

Next they look at the antecedents to unsafety. Unsafety antecedents are the situations, issues, aspects and factors of operation and activities which existed before the ‘unsafe act or condition’.

Discussing the findings, these results indicate “both complexities and subjectivity within the reporting process, and in certain case an underlying desire to apportion blame”.

This raises questions of the motivations for creating reports, seemingly “to point fingers, particularly at those in authority, or to genuinely attempt to improve conditions”.

Expectedly, categorisation of unsafe acts and conditions “was found to be highly subjective, and likely dependent first on a robust definition of what constitutes an ‘act’ and what a ‘condition’; and secondly on individuals’ interpretation of this definition”.

While some authors indicate an organisational drive to focus on individual error and blame, this dataset indicates that while human error was clearly indicated in many reports, “most unsafe acts were categorised as systemic conditions”.

Hence, “If such labels are to be used then clearer and objective definitions are needed for consistency of reporting, to mitigate the subjective nature of the process”.

Moreover, these findings suggest that we should, predictably, be suspect on statistics based on these types of taxonomies, e.g. 80% of incidents due to unsafe acts and the like, given their subjective and inconsistent categorical nature

[NB. This caution applies to most taxonomies and descriptors, e.g. what you look for is what you find.]

Link in comments.

Authors: Smith, S, Sherratt, F & Oswald, D 2017. Procurement and Law.