Analysis of Competing Hypothesis (ACH)

Analysis of Competing Hypothesis (ACH) is an intelligence analysis technique developed in the 1980s by Richard Heurer (1999) when he was an analyst at the Central Intelligence Agency. The purpose of ACH is to allow an analyst to compare all potential hypotheses against the available evidence in order to identify the most likely option among those presented.

As Heurer notes, it is a common analytical strategy to pick an explanation that seems to fit the available evidence, compare it quickly and decide either a) it fits or b) it doesn’t quite fit, and compare another hypothesis. This procedure continues by one by one, until the analyst discovers a strategy they’ll stick with.

Unfortunately, this means they may miss out on other strategies that are better explanations for the phenomenon they are seeing and therefore provide a less accurate analysis than they could otherwise generate. Additionally, this “one-by-one” approach ignores the potential for one piece of evidence to be indicative of many situations.

As an example, an employee misses work on the day after a theft. This could be explained by them being involved in the crime, having an alibi against the crime (if they can prove they were at home sick), or some other explanation. By explicitly assuming their absence makes them related, we miss other potential options.

By looking at all available hypotheses against all available evidence at once, in a grid or matrix in order to see which satisfies the most conditions.

Steps for Analysis of Competing Hypothesis

The following summary comes from Heurer’s book “Psychology of Intelligence Analysis.”

1. Brainstorm and identify hypotheses

This first step in the analysis process (which assumes that the collection process has already occurred), is to develop as many hypotheses as possible to explain the situation at hand. This is most effective if you have a group of individuals brainstorm potential explanations.

These individuals do not have to be other analysts, and in fact the brainstorming process can be helped by involving people from different areas, who may have alternate perspectives. () Hypotheses must be unique, meaning that there can be no overlapping explanations for the event in question.

In order to eliminate hypotheses later, you need to separate an unproven hypothesis (which may have no evidence for or against) from a disproven hypothesis which has confirmed evidence that it is unable to be correct.

2. Develop evidence and arguments for and against each hypothesis

While examining available evidence, it is important to note the difference between actual, concrete evidence that you have (e.g. evidence recovered from a crime scene or actions you’ve witnessed), and assumptions about, as Heurer notes “another person’s or group’s or country’s behavior, capabilities, intentions, goals, or standard operating procedures.”

Looking at each hypothesis, you should ask yourself what evidence is required to positively confirm a hypothesis or to positively disconfirm one. Do not overlook the absence of evidence which may be as indicative as the presence of evidence.

3. Assess (in)consistency of each argument or piece of evidence against each hypothesis

This can be done by hand, by placing the evidence in a matrix or table against the hypothesis (see below) or in a computerized-format.

Hypothesis 1 Hypothesis 2
Piece of evidence 1 X X
Piece of evidence 2 X

As you can see here, the first piece of evidence could be explained by hypothesis 1 or 2 while the second piece of evidence only makes sense is hypothesis 2 is implicated. Computerized tool (see the next section) can help you manage this process, including taking Credibility (the likelihood that the information is accurate) and Relevance (the likelihood that the evidence impacts the chosen hypothesis.)

4. Identify gaps in evidence

Reviewing your initial list of hypotheses again, you can now reword or refine the wording, check your Credibility and Relevance ratings, consult with other analysts to determine whether your examination of the evidence is correct, whether you’ve unfairly excluded or included evidence, and so on.

ACH software allows you to sort by “Diagnosticity”, allowing the hypotheses that are most explained by the evidence to bubble up to the top (taking into account the factors that could only be present if certain hypotheses were correct and that other factors automatically rule out certain hypotheses.)

5. Compare conclusions against software-generated options

If using the ACH software, this is where you’ll compare your software-generated Inconsistency/Weighted Inconsistency Scores (which are generated based on your consistency ratings in the previous sections) to determine if there are differences between what you believe is the inconsistency and the computer-generated inconsistency. If there is, this could be the result of a data entry error or misunderstanding the computer interface.

Attempt to disprove hypotheses in this step, so that you will be left with only hypotheses left that are likely to be correct.

If there is general agreement between your analysis and the computer-generated one, proceed to the sensitivity analysis.

6. Complete sensitivity analysis

A sensitivity analysis involves looking at what elements you consider most important to the analysis. What if you made a mistake or that evidence was wrong? Examine whether you have weighted any evidence too much or if there are alternative explanations for these items.

Perform any additional research needed to be as certain in your assumptions and evidence as you can be at this point.

7. Report conclusions

Once your ACH analysis is complete it’s important to report this. You can use standard intelligence analysis report writing.

8. Identify indicators for future

The final step is to determine indicators that will demonstrate whether your analysis was correct. In a criminal intelligence situation, you hypothesize that the failure to arrest a key drug leader in your region will result in his expanding territory and a rise in gang shootings. If you select an increase in gang murders (in the absence of a similar increase in non-gang murders) as your indicator, you’ll be able to tell if you’re on the right track.

Computerized Tools for ACH

Richard Heurer himself developed software which is now owned by Pherson Associates LLC. You can download this software here.

Additional Reading

Leave a Reply

Your email address will not be published.