CADDIS Volume 1: Stressor Identification
Causal Assessment Background
A Conceptual and Historical Explanation of Our Causal Approach
When developing the Stressor Identification guidance and CADDIS, we (Suter, Norton and Cormier) realized that causation is a difficult and controversial concept and that our methodology needed a strong conceptual foundation if it was to be useful and defensible. This module begins with a summary of the conceptual basis for CADDIS. That conceptual basis is derived from the results of our review of the history of causation and of important causal concepts. CADDIS users need not read this module, but it serves three purposes:
- It provides a means for users to better understand why the CADDIS method is what it is.
- It provides a background for anyone who wishes to modify the method for their own context.
- It presents alternative causal analysis methods proposed by others.
The Conceptual Basis for CADDIS
How can environmental assessors and managers determine the causes of environmental impairments? The first difficulty in answering this question arises because ecosystems are complex and environmental evidence is diverse. Most of the fundamental work in causal analysis has addressed simple situations having only a single candidate cause. Second, CADDIS addresses causation for specific individual cases. Most published methods for causaal analysis deal with general causation (e.g., does TCE cause childhood leukemia or does silt cause reduced caddisfly abundance) rather than specific causation (e.g., did TCE cause the leukemia cluster in Woburn, Massachusetts, or did silt cause the reduced mayfly abundance in the Willimantic River). Third, no single analytical or inferential method can accommodate the range and diversity of evidence available for cases of impairment.
A Philosophical System
Our strategy draws upon a general system of philosophy: pragmatism. Charles Sanders Peirce and his more famous followers, William James and John Dewey, developed pragmatism, which is a school of philosophy based on the premise that thinking is for doing. In other words, logic should lead to doing the right thing—the action that results in the desired outcome. To expand the utility of logic, Peirce added to deduction and induction a third type of inference, abduction (Hacking 2001, Josephson and Josephson 1996). Rather than attempting to prove that a hypothesis is correct or incorrect, abductive inference identifies the hypothesis that best explains the available information. In our case abduction determines which of the candidate causes that best accounts for the observed effects. Further, Peirce wrote that abductive inferences should be followed by deduction of the consequences of acting on the abduction and finally induction from subsequent observations should be used to support or refute the deductive predictions. For example, if we used abduction to determine that an effluent is the most likely cause, we can deduce that shutting down that source would result in recovery of the biotic community. Monitoring of water quality and biological responses could then be used inductively to infer that the effluent was indeed the cause (see the Willimantic case study). Effectively, this is adaptive management (Holling 1978, Walters 1986). Finally, Peirce believed that no method reliably delivers truth but science can provide useful approximations of the truth. Our general approach to causal assessment is pragmatic; it connects logic with action.
We began by accepting causation as a logically primitive concept (i.e., it is accepted without proof or derivation). This is justified pragmatically because we could accomplish nothing without a causal relationship to be manipulated. However, it is also justified by the realization, formalized by Hume and Kant, that humans inherently think causally without any experience, logic, or training. Ruse argued that causal explanation is an epigenetic rule, which means that causation is true in the sense that those potential ancestors who thought causally had a selective advantage.
Our Approach to Causal Inference
Our approach is an example of causal pluralism in that we accept multiple concepts of causation and all relevant evidence and methods for turning data into evidence.
Comparison of candidate causes: Although we can never prove a cause and can seldom disprove a cause, we can apply abductive inference to determine which causal hypothesis is best supported by the evidence. Hence, after defining the case (Step 1), we begin the inferential process by listing the plausible candidate causes (Step 2).
Weighing of evidence: We believe that all relevant evidence should be considered. Evidence comes from diverse sources of information such as observations at the site, regional monitoring studies, environmental manipulations, laboratory experiments, and general scientific knowledge. Information may come from the literature or may be generated ad hoc. Evidence may be generated from information by various methods including interpretation of reported observations, summary statistics, statistical modeling, and mathematical modeling. The modern tradition of weighing evidence of causation is based on Hill's “criteria.” However, to provide some transparency and consistency, CADDIS adopts the scoring system developed by Susser and introduced to ecologists by Fox. Weighing evidence requires that the evidence be categorized. CADDIS uses 17 types of evidence from the site (Step 3) and from elsewhere (Step 4). The scores represent the relevance and quality of evidence provided by different outcomes for each type of evidence.
Rejection: Like Popper, we recognize that one can more confidently eliminate a causal hypothesis than accept it. For example, if an effect occurs downstream of a source, that is weak supporting evidence for emissions from that source as a cause, but, if the effect occurs upstream of the source, rejection with confidence is possible. Therefore, we begin the weighing of evidence by rejecting as many causes as possible. However, rejection is not sufficient because we can never reject all but one candidate cause. Rejection requires evidence from the site and is expressed as an R score in the Strength-of-Evidence tables (Step 3) and is sufficient to negate all positive evidence for a candidate cause.
Diagnosis: Diagnosis is the determination of a cause based on characteristics or aspects of the effects (i.e., symptoms such as eroded fins or “cigar burn” lesions or accumulation of a chemical in certain organs). In the environmental sciences, it is employed by pathologists in investigations of fish and wildlife kills, although, even there, other types of evidence also are used. Community-level diagnostics have been attempted in ecological research, but, to date, their application and reliability have been very limited. In CADDIS, diagnosis is treated as an extreme case of symptomatic evidence and is given a D score. A high-quality diagnosis is sufficient to negate all other evidence for a candidate cause.
Synthesis of evidence: After each candidate cause is evaluated with respect to all available types of evidence, assessors must compare the alternative candidate causes. In many cases, one candidate cause will be clearly more consistent with the evidence. If not, the assessors should consider the potential sources of uncertainty—including lack of data, poor quality data, a poorly defined impairment, and multiple causes. In difficult cases, condensing the types of evidence to a few characteristics may be helpful (Step 5). If the assessors have identified the most likely cause or causes with sufficient confidence, remedial actions may be planned and carried out.
Summary: The essence of the CADDIS approach to causal inference is the comparison of alternate candidate causes by determining which is best supported by the totality of evidence. Its standard process provides transparency and reduces inferential errors without restricting the types of evidence used.