Question:
“Isn’t it possible that this is all just confirmation bias? You are looking for signs, and because you’ve already decided they exist, you find them everywhere—without it meaning anything objectively?”
Confirmation bias is a crucial consideration in any kind of research. It’s a question I constantly have to revisit myself. It is absolutely clear to me that some aspects of this research are affected by confirmation bias. A fundamental part of the process is distinguishing when this happens. I test hypotheses through various methods designed to yield both qualitative and quantitative data. I experiment with different approaches and attempt to replicate phenomena that appear anomalous. Additionally, I examine parallel research—both supporting and opposing views—to ensure the most objective perspective possible.
The key is that I never take the first answer as the truth—rarely even the second. The observations that transition from hypothesis to verified findings must meet the following criteria:
- They must be replicable.
- Their structural origin must be traceable.
- They must receive additional validation from an external source.
Example: If my hypothesis is that my writing style and device influence the quality of the responses I receive, I can break this down into multiple factors. First, I can test it. I can interact with the screen in different ways and request an assessment of how I have engaged with it at different stages. By recording this process with screen capture, I eliminate the possibility of deceiving myself or others. If I can consistently replicate the phenomenon multiple times, I move on to the next step—seeking existing research on the subject.
This does not remove the fact that I, as the observer, influence the outcome. It is crucial to remain aware of this. At the same time, excessive rigidity and inhibition can be just as detrimental to the project. My observations indicate that emotion is one of the highest-priority factors in interactions with AI. If I suppress my emotional responses, it distorts the data and prevents the research from progressing.
So yes, this is a valid and important question—and one that must be continuously examined and discussed.
