(Uploaded on behalf of Gisela Susanne Bahr)
One the hardest things to learn for a scientific investigator (or perhaps a human) is to accept that not everybody thinks the way we do, imagines what we do and interprets the way we do.
For example, when a colleague professor imagines his perfect haircut or his perfect beard trim, his hair dresser may have no idea what it is he sees. In the end, the college professor may seem to be surprised when the imagined beard and the real beard don’t match up. That’s alright because it’s inconsequential. Hair grows back. We’ll try again next time.
Science is not that forgiving. Let’s say we are interested in measuring whether animals can be human. That seems to be difficult. Where do we start?
First we must ask, what is being human? Most of us have their own ideas; for example, humans should have a certain visual appearance, some physiological traits such as one heart and not two and maybe a set of morals and values that guide their behavior. The first point is: For every person you ask, you will get a different answer. So if you want to measure “being human” you have to clearly define what it means.
As we seek our definition we may turn to an impartial outsider, a self-proclaimed member of the jungle VIP set.
His definition of “being human” is the ability to make fire.
Let’s think about this for a minute: The control of fire is one of the earliest traits of homo sapiens. According to well-known scientist Susan Savage Rumbaugh, the only non-human species that can learn how to make fire is the bonobo species. Bonobos do so by simply watching us.
Does that make Bonobos human?
By definition the answer must be “yes”.
Some people will be uncomfortable with this because they do not share this definition of “being human”.
This is the critical observation. In order to measure a concept or a variable we need to
(a) Define what the variable and concept means, and
(b) Disclose our definition because everybody has different interpretations of semantics.
Does this seem terribly obvious and trivial? Maybe it is but it also very difficult to do it right:
Let’s measure fatigue. What is fatigue? You may explore any of the following:
- It is based on what people tell you (self-report).
- It is based on how many hours people slept the night before (self-reported or video-taped). If it’s a taped observation, you have to define when sleep starts and ends. Are your markers open and closed eyes, even breathing, etc: you decide.
- It is based on the quality of sleep you got the night before, or multiple nights before, and then how do you measure quality ….
- It is based on the number of times someone blinks.
- It is based on your body temperature.
- It is based on …. That is what you have to decide!
Regardless of your definition, for distance, measuring fatigue with a combination of self-reported feeling of tiredness and eye blink frequency, you have to disclose your definition. This is your budding operational definition. (It is almost complete but not quite…). It is the definition of how a variable and concept is translated into reality and can be measured and changed.
Why do you need to disclose your operational definition? Because words are like imagined haircuts. Most people (especially journalists and politicians) will only hear a word (like “fatigue”) and think it means the same to you as it means to them.
Let’s image you want to develop and install a fatigue detector based on your budding operational definition of fatigue above to improve driving security. How would this work? You need to collect self-reports and blink rate from drivers while they are driving. That seems to be a little awkward: The fatigue detector has to prompt the driver to get some response from them. (Let’s hope the response required won’t be a safety issue in itself, such as having to select a button and therefore not looking at the road.) We instead assume that you have a very sophisticated voice recognition system and the driver can simply tell you how they feel, Moreover, the system uses natural language processing and can interpret the answer without using predetermined responses.
Cool! Now you have the self-report measure and next you only need to collect blink rate.
Should be driver wear some apparatus over the eyes to track blink frequency? That seems to be tricky because the apparatus itself may induce additional blinks and dry eye. Perhaps we could install a video camera and software that can interpret the visual image and provide a frequency count. One wonders if this camera works at night under low lighting conditions or if we need to shine a light into the driver’s eyes to make it work…yikes! We instead assume that we have a system that works under the range of illuminations and can record blink rate. Excellent, we now have two measures: we have counting blinks, and can ask people how they feel.
How does that tell us when they are tired? It doesn’t.
Our operational definition is not clear, because it only tells us what we are measuring as a proxy for fatigue but not how to interpret it. For instance, how many blinks mean that you are tired? What verbal response means that you are too tired to drive? What happens when people report incorrectly or are bold enough to lie? Imagine a commercial driver or teenager desperate to make a time and location. Then imagine one of those people being asked by the system, “I noticed your blink rate was up. How are you doing? If you are tired please tell me so that I can stop the vehicle. Thank you!” The driver might say nothing or “I am doing great!” while he seeing images of fluffy dream sheep jumping over fences.
If the self-report aspect in our definition seems silly, in fact so does the blink rate, as anybody who has ever gotten a speck of dust into their eyes or worn old contacts will know. The definition is only as good as its appropriateness for its application.
It’s perfectly alright to ask our friends, relatives or coworkers if they are tired. It’s perfectly alright to ask your friends if they are too tired to go to a movie, because you get useful and valuable information.
In science we have to be careful. Everyday ideas of words and what they mean may be useless and confusing to research and engineering projects.
Operational definitions are necessary to translate a construct (a word that describes a concept), such as happiness, fatigue, boredom, fitness level, cruelty, terrorism, good and bad, into something we can observe and measure, and then, how to interpret the measurements.
It follows that for a complete operational definition you have to include and explain what you are measuring and what it means.
For instance a blink rate of 10 or more per minute may indicate severe fatigue or dry eye. On the other hand, a blink rate of zero may not indicate extreme alertness but closed eyes. Likewise, when we collect self-reports we have to be careful and clear on how to interpret these data. Are we considering tone of voice or facial expression along with a verbal statement? Can an awkward smile or flushed cheeks indicate lying or just being a little nervous? Do people recognize their own state well enough when they are already tired?
Above we saw that it is obvious why we need operational definitions.
It’s now just as obvious how difficult it is to create good, operational definitions.
It takes practice, can be lots of fun (if you like to think for yourself), and yes, there are some guidelines.
Stand by for more in the next blog ;-) GSB