The Mattson Consensus

At the start of a recent search, our SAR Coordinator handed each of us at the staging area a piece of paper, and on that paper was written the following:

ROW=                        A-I

1
2
3
4
5
6
7
8
9
10

I stared at him blankly, knowing only that ROW stands for "Rest of the World," and then he explained:

The numbers one through ten corresponded with the ten numbered segments he'd drawn on the map laid out on the hood of his vehicle. First, with each of us working independently without discussing the values, we'd give a percentage to the ROW. In other words, what's the probability the missing subject is not in the search area--segments 1-10--at all? And, regardless of what some of us believed, we could not assign the ROW a probability of 100%. We had to leave some room for the search area. So I estimated generously low and wrote 70% on my paper.

Next, we'd assign each numbered segment a lettered value, A through I, as follows:

A - very likely in this segment
B
C - likely in this segment
D
E - even chance
F
G - unlikely in this segment
H
I - very unlikely in this segment

I looked from the map to my paper and back time and time again and kept changing my mind, second-guessing my assumptions. I kept having to remind myself that the presumption was supposed to be a "despondent subject" as opposed to, let's say, "stranded subject trying to get home."

After I and the rest of the group of searchers, deputies and detectives filled out our papers, we handed them to the Coordinator, who then turned to his laptop and started punching in numbers as he went through each page.

I had no idea at the time that what he was doing was using what's called the "Mattson Consensus," with an experimental lettering method developed by Dan O'Connor, a former contract helicopter pilot for flight operations at Grand Canyon National Park.

Typically, the Mattson Consensus employs the use of percentages for each search segment. In the end, the sum of each "expert's" percentages, including that of the ROW and all designated search segments, must total 100%, with no segment being assigned 0%. (A zero would mean that the expert knew for a fact that the subject was not in a particular area.)

According to Mattson, it's best to perform this exercise privately, "because it will ensure that even meeker individuals will be able to express their opinion without being intimidated by the more vocal members of the group." (See: Search & Rescue and The Wisdom of Crowds)

With Dan O'Connor's method, the letters A through I are used for each search segment instead of percentages, and then a numeric value is assigned to each letter. This is what our SAR Coordinator was doing when he was entering all of our results into his CASIE computer program, resulting in a plan to first search the segments with the highest probability of area (POA) as suggested by our group's consensus. The computer figures out the segment percentages, rather than the members of the group having to do so.

The following explanation of the numeric values assigned to each letter is quoted from the "CASIE Help: Planning" section of the math.arizona.edu website:

Each letter that the expert has used is assigned a numerical value according to the scheme:

A = 9, B = 8, ..., I = 1, if the lowest letter used by that expert is an I.
A = 8, B = 7, ..., H = 1, if the lowest letter used by that expert is an H.
A = 7, B = 6, ..., G = 1, if the lowest letter used by that expert is an G.
A = 6, B = 5, ..., F = 1, if the lowest letter used by that expert is an F.
A = 5, B = 4, ..., E = 1, if the lowest letter used by that expert is an E.
A = 4, B = 3, ..., D = 1, if the lowest letter used by that expert is an D.
A = 3, B = 2, C = 1, if the lowest letter used by that expert is an C.
A = 2, B = 1, if the lowest letter used by that expert is an B.
A = 1, if the lowest letter used by that expert is an A.

Now the expert's total is obtained, and the ratio of the expert's numerically assigned value to the expert's total is that expert's POA for that segment.

An example may clear up any confusion. Imagine that an expert assigns a G to segment 1, an A to segment 2, and a G to the R.O.W. The lowest letter is a G, so we use the third line of the above table. The expert's total will be 9 (7 for the A, and 1 for each G). This expert's POA for segment 1 is 1/9, for segment 2 is 7/9, and for the R.O.W. is 1/9.

For our team's search, the resulting average percentage for the ROW was 54%, with segments 2, 3 and 9 having the highest POA. I don't yet have an answer as to the accuracy of this exercise in this particular search, because it is still ongoing, but one example of a real-life mission that used the Mattson Consensus is that involving Ranger Randy Morgenson, who went missing while on a solo backcountry patrol in the Sequoia and Kings Canyon National Park, California. While searchers did fail to find Ranger Morgenson's body during the mission period (eventually determined to be due to the high amount of runoff that summer), he was ultimately found 5 years later "within an area of high probability of discovery in the original search." Read the story here.