Wandering Out Into The Deep End...
As usual (and expected), my immediate environment and daily social interactions have lent me the tools to engage in deep thought about my own field of science. My line of thinking this week however, begins and ends in the same place, despite its rigorous journey through UNT Department of Behavior Analysis. About two weeks ago, during one of my weekly meetings, a discussion was started on involving a research component to the online education program. The discussion progressed into the use of available technologies to teach the core concepts of behavior analysis, such as guided virtual environments and emerging development platforms (i.e., Ardiuno, Spark Core, etc.). We not only discussed how these technologies can be used to teach the students, but also how we can use them to inform future directions for the online program.
More recently, while conversing about some of my current progress in BAO with the lead web developer, he made an off-the-cuff statement about ”…trying to reach the future.” Impromptu as it may have been, this was enough to thread a string through the snippets of conversation and lectures that I had partaken in though out the week. More excitingly however, it eventually provided a train thought that looped me back to various portions of previous posts.
I have mentioned before that behavior analysis as an independent science can benefit greatly from moving forward with modern technology. I have started to think about the role stimulus discrimination, hierarchal stimulus relations, and the feedback loops that regulate the future probability of behavior based on an organism’s history. As developing programs for various tasks, and BAO are a common routine for me, it seems inevitable that curiosities regarding artificial intelligence (AI) have started to crop up. I have started to ask myself what the flaws have been in the current development of AI programs and how they can be improved. One of my professors man a comment intended to be nothing more than a remark about cognitive psychology and it’s flaws stuck with me. He said that the reason why most artificial intelligence programs fail is because they focus on dimensional qualities of stimuli (or what they look like) as opposed to the rule governed qualities of stimuli (or behaviors emitted by individuals based on self talk or behaviors that have been previous reinforced by their verbal community).
The idea of governing behavior based on rules that have had reinforcing and/or punishing consequences seems more logical than writing code that attempts to describe dimensions such as upside down or inside out. This now returns us back to our initial discussion in the first meeting. The discussion of including a research component into BAO would necessarily involve a style of experimentation that has become a tradition in applied behavior analysis: changing future procedures based on previous data. In other words, if one is trying to progress with a patient in a clinic, it is likely more useful, and sometimes even safer to change ones procedures if the current procedures are having neutral or deleterious effects on behavior. In other words, future changes are predicated on previous data, and behavior analysts are if nothing if not obsessed with collecting data any way they can. This definition fits perfectly with a concept that is core to artificial intelligence, known as machine learning. Machine learning is facet of computer science in which computer programs are designed to make future changes based on collected data.
How does this relate to BAO? Skinner’s teaching machines were designed to cater to the individual needs of each student so that education can be an efficient and thorough process. This provides BAO with a footing to conduct possibly very interesting experiments in with regards to online education. Comparisons can be made between students that are educated with courses designed to change based on data from their previously completed assignments, with both static (or course work that does not change) courses that only move forward, and a the more common web-based course which simply allows the user to attempt each lesson until they have achieved mastery (Skinner’s original teaching machine concept).
Reaching the end of this thought, I must make an author’s note. I am aware that the ideas presented here may seem both unusual and perhaps even borderline crazy, but they are ideas none-the-less. The thing about ideas, which I have come to appreciate about these commentaries, is that they often at worst breed new ideas and questions, but at best, develop into fruitful lines of research. It is likely apparent by now that I hold the goals of science high on my list of priorities. It is my philosophy that the best way to understand is to tread out into the waters of curiosity and splash around a bit.
