The Computer Is Objective?
We've all heard the old saw, "The analysis must be objective because a computer did it." Nothing could be further from the truth.
There are two elements to every computer: hardware and software. Hardware is the computer circuitry, and is objective. But hardware is useless (except as a very accurate clock) without software. Software - programs - instruct the hardware as to what to do. Software is designed and written by people. And we know that people can be less than objective, can have personal agendas.
Allow me to present four examples that support this point.
First, let's examine the current situation in college football. The College Football Playoff selection committee, which currently has Alabama ranked Number 1, is comprised of 13 people, including one who didn't even play college football.
So, in an effort to be objective, we turn to computers. One computer program that ranks college football teams is the Anderson & Hester Rankings. It claims to be "distinct in four ways:"
- "Unlike the polls, these rankings do not reward teams for running up scores. Teams are rewarded for beating quality opponents..."What is a quality opponent? No definition is given.
- "Unlike the polls, these rankings do not prejudge teams... each team's ranking reflects its actual accomplishments..."What constitutes actual accomplishments?Again, no definition is given.
- "These rankings compute the most accurate strength of schedule ratings.Each team's opponents and opponents' opponents are judged not only by their won-lost records but also, uniquely, by their conferences' strength." Key word here: judged.And no definition of strength is given.
- "These rankings provide the most accurate conference ratings. Each conference is rated according to its non-conference won-lost record and the difficulty of its non-conference schedule." What does 'difficulty of its non-conference schedule' mean? Can that be non-objectively manipulated?
I am not trying to destroy Anderson & Hester's ranking program. I am simply pointing out that the ranking program, even though it’s run on a computer, was written by people, and their non-objective biases could have been written into it. I have identified four factors that the program uses that are not at all objective. Plus Anderson and Hester are free to enter any non-objective factor values they want.
In the interest of full disclosure, I graduated from FSU, I played (rode the bench, not at FSU) college football, and was a computer programmer. So I'm not some know-nothing loudmouth. And I should love Anderson & Hester's rankings that place FSU at #1. Am I being objective?
Now let's turn our attention to politics. One of my favorites is a program called a "Structural Topic Model" (STM). The STM program itself may be as objective as its programmers could make it, but its use is subject to non-objective influences.
The STM program is comprised of numerous objective mathematical formulae, such as:
Conditional on the topic chosen, draw an observed word from that topic.
ωd,n в€Ј χd,n, βd, k=z, ˜ MultinomialI(βd,k=z)
The last phrase: "Conditional on the topic chosen ..." is the tipoff. The entire process is driven by the topic chosen, which may or may not be objective. If the topic is not objectively chosen, then the entire analysis is not objective. The computer and program themselves may be objective, but if the use is not objective, the entire analysis is subjective -- period.
Downright unobjective (and agenda driven) is PolicyMaker 4, a program that says it "...is a rapid assessment method for analyzing and managing the politics of public policy" and manages "what gets on the agenda..." The fact that it's run on a computer doesn't make it objective.
PolicyMaker is a rapid assessment method for analyzing and managing the politics of public policy. Politics affects all aspects of public policy -- what gets on the agenda, who supports an issue, who opposes an issue, whether an issue receives official approval, and whether the official policy is implemented. PolicyMaker is a logical and formal procedure to provide practical advice on how to manage the political aspects of public policy. The method helps decision-makers improve the political feasibility of their policy.
Let me guess. What the program highlights on the agenda for backer notification and therefore turnout are issues that its developers and users support. People wrote PolicyMaker 4, and it reflects their biases.
One PolicyMaker 4 user, Dr. Michael Reich, has this item on his website: a link to a webpage entitled “Getting to universal health care”. I'm suuure Dr. Reich makes objective use of PolicyMaker 4.
Then there's the ultimate of objectivity (NOT), the "Elements of Family Planning Success Toolkit." Its website has this:
K4Health's predecessor (The INFO Project) initiated the Elements of Family Planning Success activity to provide recommendations made by program managers and policy makers to program managers and policy makers. Health care professionals from around the world identified the top 10 elements essential for the success of family planning programs. The INFO Project and the follow-on K4Health Project then merged this local knowledge with evidence-based information gathered worldwide to develop a package of resources, including this toolkit. [emphasis in the original]
There is one sentence that permits the questioning of the objectivity of the entire toolkit (program): "Health care professionals from around the world identified the top 10 elements essential for the success of family planning programs." The K4Health people didn't identify what experts consider to be essential elements just to have something to do. They used the elements in development of the program. Are we then surprised when the program recommends family planning?
Is the “Sex Check! Are You Ready For Sex?” an example of the “objective” computer program's use by Planned Parenthood?
Computer analysis presents a conclusion that is based entirely upon analysis done by software.
Yes, conservatives can do exactly the same thing. So the next time you read an article that claims objectivity because its findings are based upon computer analysis, you have the knowledge necessary to dismiss it and its author(s). Computers are great for tabulation, but analysis will always be less than objective because it is done by software. And software is written by non-objective people. And computers don't help. Just ask a Baylor or T.C.U. fan.
No opinion, just facts.
Dr. Warren Beatty (not the liberal actor) earned a Ph.D. in quantitative management and statistics from Florida State University. He was a (very conservative) professor of quantitative management specializing in using statistics to assist/support decision-making. He has been a consultant to many small businesses and is now retired. Dr. Beatty is a veteran who served in the U.S. Army for 22 years. He blogs at rwno.limewebs.com.