The Assumptions We Bring to Fire Scenes: Confirmation Bias in Origin and Cause Investigation

Adapted from a presentation on Confirmation Bias delivered by Vithyaa Thavapalan at the Sydney August Gathering of Forensic Engineering Society Australia One of the biggest risks in forensic work is not always what we miss at a scene. Sometimes, it is what we think we already know.

In fire origin and cause investigation, early information can be powerful. A witness may mention an electrical issue. A property owner may refer to a faulty appliance. Police may advise that a matter is being treated as suspicious. An insurer may provide a theory in the referral. A heavily damaged area may appear to draw immediate attention. Each of these pieces of information may be relevant, but none of them should become the conclusion before the evidence has been properly examined.

This is where confirmation bias can enter the process. Confirmation bias is the tendency to seek, interpret, or give greater weight to information that supports an existing belief or working theory, while overlooking or minimising information that challenges it. In forensic investigation, this does not usually happen because someone is careless or intentionally biased. It often happens because investigators are human. We are trained to find patterns, form explanations, and make sense of incomplete information.

The issue is not whether investigators form early hypotheses. In many cases, we have to. A working hypothesis helps direct the investigation, identify what evidence needs to be collected, and determine what questions need to be answered. The risk arises when that early hypothesis becomes fixed too soon, and the investigation starts to look for evidence that confirms it rather than evidence that tests it.

Fire scenes are rarely neat. They are dynamic, damaged, altered by fire suppression, affected by ventilation, and often examined after people have already entered, secured, disturbed, or discussed the scene. By the time an investigator arrives, they may already have been given a version of events. That version may come from an owner, occupant, witness, fire service, police, insurer, loss adjuster, builder, electrician, or another expert. The information may be useful, but it can also quietly frame the way the scene is viewed.

A known appliance issue may cause an investigator to focus too quickly on mechanical or electrical failure. A lack of obvious forced entry may cause an investigator to consider internal involvement before other possibilities have been properly assessed. A suspicious circumstance may cause ordinary fire damage to be viewed through the lens of deliberate ignition. A heavily damaged area may be mistaken for the area of origin without adequate consideration of fire spread, ventilation, fuel load, suppression effects, or post-fire disturbance.

The danger is not the information itself. The danger is allowing that information to become the explanation before it has been tested.

This is particularly important when considering electrical evidence. A fire scene may contain damaged wiring, damaged appliances, tripped circuits, melted conductors, or electrical components within an area of damage. However, the presence of electrical damage does not automatically mean the fire was caused by electrical energy. Electrical systems can be damaged by fire, just as they can be involved in causing fire. The investigator must consider whether the electrical evidence supports ignition, or whether it is better explained as damage resulting from exposure to an external fire.

The same caution applies to fire patterns. Fire patterns can provide important information about fire development and areas of interest, but they must be interpreted in context. The area of greatest damage is not automatically the area of origin. Irregular patterns are not automatically evidence of ignitable liquid. Isolated damage is not automatically deliberate. Each observation needs to be considered alongside the broader scene, fuel packages, ventilation, witness information, fire service activity, laboratory results, and any other available data.

A defensible origin and cause opinion requires investigators to separate what was observed, what was reported, what was tested, what was inferred, and what could not be determined. This distinction is one of the strongest safeguards against assumption-based reasoning. It also assists the reader, whether that reader is an insurer, lawyer, engineer, court, tribunal, or another expert, to understand how the opinion was reached.

Suspicious circumstances are also not a substitute for origin and cause evidence. A missing item, unusual behaviour, inconsistent accounts, or police interest may all be relevant to the broader investigation, but they do not replace the need to determine where the fire originated, how it developed, and what ignition source and first fuel package are supported by the physical evidence. In the same way, an accidental explanation should not be accepted simply because a possible appliance or electrical item exists in the area. Possibility is not the same as probability, and probability still needs to be supported by evidence.

This is why competing hypotheses matter. A good origin and cause investigation does not set out to prove the first theory. It develops and tests alternative explanations. What evidence supports this hypothesis? What evidence challenges it? What other explanations remain possible? What evidence would cause the investigator to change their opinion? Has possibility been confused with probability? Have cause and responsibility been kept separate?

That final question is important. Determining the cause of a fire is not always the same as determining responsibility for the fire. An investigator may be able to conclude that a fire was deliberately lit, but that does not necessarily identify who lit it. Similarly, an investigator may identify an area of origin without being able to determine a precise ignition source. Clear reporting requires clear boundaries around what can and cannot be concluded.

Confirmation bias cannot be managed by simply saying, “I am objective.” Objectivity has to be built into the process. One practical safeguard is the use of a hypothesis table, where each plausible hypothesis is recorded alongside the evidence that supports and challenges it. This helps prevent alternatives from being silently discarded and allows the reasoning process to be followed.

Another safeguard is to build in a bias checkpoint during the analysis. Investigators can ask themselves: what was my first theory, what made that theory attractive, what evidence would disprove it, and have I actively looked for evidence that challenges it? This is particularly useful where the referral, witness information, scene observations, or stakeholder expectations point strongly in one direction.

It is also important to separate observation from interpretation in reports. For example, “the wiring was fire damaged” is an observation. “The wiring caused the fire” is an interpretation that requires supporting evidence. “The appliance was located in the area of damage” is an observation. “The appliance was the ignition source” is an opinion that must be tested against the physical evidence and competing hypotheses.

The language used in expert reporting matters. A strong report does not overreach. It explains what the evidence supports, what has been eliminated, what remains uncertain, and what limitations apply. Phrases such as “supports,” “is consistent with,” “does not support,” and “cannot be eliminated” can be useful when they are used carefully and accurately. The goal is not to sound more certain than the evidence allows. The goal is to communicate the opinion with clarity, transparency, and proper evidentiary support.

Forensic investigation is not just about reaching an answer. It is about being able to explain how that answer was reached. A defensible opinion should show the path from observation to interpretation to conclusion. It should identify the evidence relied upon, explain why alternatives were rejected, and acknowledge any limitations.

Confirmation bias is not avoided by pretending we do not form assumptions. We all do. It is avoided by making those assumptions visible, testing them against the evidence, and being willing to change our opinion when the evidence requires it.

In origin and cause investigation, that discipline matters. Because the question is not only, “what do we think happened?” The better question is, “what does the evidence actually support?”