Wednesday, April 27, 2016

Thorndike and Skinner



Edward L. Thorndike
1874-1949
Burrhus F. Skinner 
(1904-1990)
Two of the giants in the behaviorist tradition of learning theory were Edward Thorndike and B.F. Skinner. Skinner took Thorndike's work on behavior shaping to another level of complexity. But with greater complexity comes greater potential for being misunderstood, and that's exactly what happened. 

Thorndike observed a century ago that when a behavior is followed by a pleasant effect, the behavior is likely to be repeated in the same situation. Behavior followed by an unpleasant effect is unlikely to be repeated. Thorndike's Law of Effect is easy to understand, due partly to the language he used in describing it. There is no ambiguity in what he meant by pleasant, unpleasant, or effect.

The same cannot be said for B.F. Skinner. He called these same two ways of shaping behavior using pleasant and unpleasant consequences positive reinforcement and positive punishment. Positive didn't mean that they were good or kind or effective, as you might expect; positive simply meant that the effect was added in response to the behavior. Skinner proposed that reinforcement and punishment could also be done by subtracting an effect that was already at work when the behavior was exhibited. He called this negative reinforcement and negative punishment, respectively. Skinner's four ways of shaping behavior are known as the four quadrants of Operant Conditioning. Unfortunately, three of the four are routinely misunderstood.

Positive punishment seems like an oxymoron and negative punishment seems redundant because of Skinner's naming scheme. Those terms are seldom used outside of academic circles. In contrast, positive reinforcement has slipped into common language as a synonym for reward. No harm there. That's exactly what it is. 

But negative reinforcement is problematic. Too often this term is used casually as a synonym for punishment. It's easy to understand why. Negative has many meanings in the English language but the most common connotes undesirability and unpleasantness. Skinner used negative in a mathematical sense. Negative reinforcement is a way of rewarding behavior by taking away an aversive stimulus already acting on the subject. In horse training, for example, you may exert pressure on a horse to get him to move in a certain way, then instantly remove the pressure when he gives an acceptable try. The removal of an aversive is a reward, a pleasant effect. If you have trouble wrapping your mind around this, don't feel bad.

Not only is negative reinforcement incorrectly associated with punishment, it suffers from the association. Many people find punishing their children or animals distasteful and turn to other options. Clicker Training for dogs and horses, for example, uses a clicking sound and food treats to reward correct behavior. Incorrect behavior is ignored, or a different cue is given in hopes of getting a response that can be rewarded. Punishment is reserved for dealing with extremes of undesirable or dangerous behavior.

Learning theory is one of my favorite topics so I have studied Skinner's work many times. I'm always saddened just a bit by the language he chose because I know how often it is misunderstood. By the same token, I love the elegant simplicity of Thorndike's description of his Law of Effect. In the end, words do matter.     

Friday, April 1, 2016

Tips for Analyzing Research Articles

Identifying the Research Problem

Research articles are all about research problems. Sometimes this key piece of the puzzle is clearly identified. If not, you have to dig it out. It helps to remember that the research problem is an important gap in our knowledge about a real-world problem. For example, suppose you are interested in the attrition rate among graduate students. This is a real-world problem. It is important because it has negative repercussions for students. The fact that the reasons for attrition aren’t fully understood makes attrition a research problem. An investigation of a research problem may not solve the real-world problem, but it is usually a step in the right direction.
One helpful approach for digging out the research problem is to answer three questions about the article: What? So What? and Because Why? If you can clearly answer each question in one or two statements, you probably have a pretty good grip on the research problem, and you could probably explain it to someone else without their eyes glazing over. Let’s look at each question.
WHAT? This question focuses on the real-world problem and the specific gap in what we know about it. Your answer to the “What?” question might take the form, “Although A, not B.” For example, “Although student attrition has been shown to be an ongoing concern for many institutions, the reasons that students drop out are still not fully understood.”
SO WHAT? This question focuses on the undesirable consequences of the gap in our knowledge of the real-world problem; i.e., why anyone should care. Your answer to the “So What?” question might take the form. “Without B, C.” For example, “Without an understanding of the reasons students drop out, there is little chance of helping at-risk students reach their learning goals.”
BECAUSE WHY? This question focuses on the theoretical foundation or conceptual basis for the research problem. You don’t need to overthink this! Just point to the broad subject areas that relate. “Because Why?” statements might take the form, “This problem spans the domains of D, E, and F.” For example, “The problem of the lack in understanding of the causes of student attrition spans the domains of motivation theory, individual differences theory, and transactional distance theory.”

What a Problem Isn’t

Identifying a real-world problem might seem easy to do. But let’s think about problems for a moment. Often what seems like a problem is really just a state of affairs. For example, “It’s raining” is not a problem but a state of affairs. If you’re a farmer worried that your crops aren’t getting enough moisture, rain is a good thing. If you’re clinging to your rooftop with flood waters rising all around you, rain is not so good. The point is that context is what turns a state of affairs into a problem. For example, “It’s raining and I have to walk across campus to my next class” provides some context and gets us much closer to a real problem, but even this leaves room for interpretation. “It’s raining, I have no umbrella, I have to walk across campus to my next class, and I’m already catching a cold” leaves little room for doubt. You have a problem! 

Research Type

After you’ve nailed down the research problem, it helps to determine the type of research being done. Here’s a quick overview of how research type is traditionally broken down:
Hypothesis-testing research manipulates variables and studies causality. It may be experimental in which assignment to experimental and control groups is completely random. This is is relatively rare because completely random assignment is hard to achieve. More likely it is quasi-experimental, wherein the experiment uses a convenient population of participants and assignment to experimental and control groups in not totally random. "Experiment" in research simply means that some treatment is performed and the effects are observed.
Descriptive research describes something that already exists. It may be historical, correlational, or a case study. Problem-solving research seeks to fill needs, right wrongs, or improve situations. It may take the form of theory building, design and development, or action research.

Research Data

Research requires data. Without data of some kind, an article is an opinion piece. There’s nothing wrong with opinion pieces, but they are not research. The nature of the data, the way it is collected, and how it is analyzed is determined by the research design. Data may be numbers such as test scores, visits to a web site, or machine measurements. How people feel can also be turned into numeric data by using survey questions that ask them to rate their level of agreement or disagreement with a statement. These are called Likert-scale questions and they can be converted to numeric data suitable for quantitative analysis. (By the way, it’s pronounced LICK-urt, not LIKE-urt.) Numeric data are analyzed with software such as SPSS (Statistical Program for the Social Sciences) using quantitative methods that yield statistical results. The results allow conclusions to be reached and predictions to be made about other individual cases. Using a general rule to predict specific outcomes is deductive reasoning. 
Data don’t have to be numeric. Data can also be subjective and unstructured. For example, interviews, observations, recordings of participants thinking out loud, and open-ended survey questions are all valid research data. This data must be analyzed using qualitative methods, which seek to identify ideas, concepts, and categories that appear in textual data and reduce them through a sort of distilling process until an overarching theme is obtained. This is an inductive process; i.e., using specific outcomes to infer a general rule. Qualitative analysis is like outlining in reverse. You begin with detail and group the detail together into more general categories, then group the categories into broader descriptions still until you reach the highest level. That is when you have your overarching theme, conclusion, or rule. Data are collected with an instrument, which simply refers to the tool used to get the data. The data collection instrument may be a survey, or a test, or simply the researcher’s observations. Whatever means is used to get the data is the data collection instrument.

Showing Your Understanding

Understanding the research articles you read is one thing; putting that understanding in writing is something else entirely and is usually done as part of a literature review. The literature review might seem overwhelming so here’s a trick to break it down into manageable chunks: Do your analysis of each article as you find it.  A good format to use is the annotated bibliography citation. An AB is just what it sounds like: a bibliography or list of sources with notes about each. The AB entry starts with the proper APA citation for the article. If you use the article later, it’s a simple matter to copy and paste the citation into your list of references. Following the citation is 150 words of descriptive and evaluative text. This is a concise narrative written in your own words that identifies the problem investigated, methodology used, results and conclusions, and your opinion about what the article adds to the body of knowledge on the article’s subject. When you construct your lit review, you can draw your discussion of the article from the AB narrative you created earlier, fleshing it out as needed. Sometimes you can even use an AB entry verbatim. 

The Literature Review

A good literature review is more than a collection of article analyses, however. It is a cohesive, smooth-flowing document of its own. Thus, it needs to have a logical, hierarchical organization. It must also contain introductory, transitional, and summarizing sections that connect the ideas contained in the articles. This aspect of creating a literature review has the quality of synthesis; i.e., bringing together discrete parts to make a whole. Your finished literature review should be written clearly and economically, be grammatical, and adhere to the style and format rules that apply. Minimize the use of quotes. You might think that quoting an author makes your writing more academic or more credible. It doesn’t. Quoting can make your writing seem choppy and even lazy. It’s better to paraphrase an author’s work in your own writing style. You might even say it better! While we’re at it, beware of making your writing overly formal or academic. Before using an uncommon word, ask yourself if a common word would do. Remember, someone unfamiliar with the subject should still be able to get a basic sense of what is going on through your writing alone. The best academic writing is free of ego. Its purpose is to communicate clearly, not demonstrate the writer’s intelligence or mastery of the subject. 

And Then There’s APA

For writing in education, and the social and behavioral sciences, the style and format rules that apply are those of the American Psychological Association. The APA publication manual is a must-have. The APA rules may seem foreign, arbitrary, or even illogical to you at first. Get over it. APA is one of those things in life that you just have to accept and make the best of. The good news is that you’ll become a better and more disciplined writer as a result. You’ll also begin to take pride in your ability to write within the APA framework. Style rules are common in all industries and good writers do not feel constrained by them any more than they feel constrained by traffic laws.

Final Word

Even when they follow the rules, research articles are seldom easy to read. Take heart. You’ll get familiar with the structure and the language used.  You’ll get faster at forming an impression about an article, and if you choose to use it, you’ll know instinctively how to describe it. And of course, you will learn what not to do in your own writing. There is even a chance that you’ll come to enjoy research. Imagine that!