Even when we take every precaution to make sure that the biases of researchers or subjects don’t come into play through blinding, it’s possible to introduce unwanted variation into experiments through just doing research itself. Observer effects are important, and need to be considered in order to ensure that research is reproducible. They’re the topic of this module in Experimental Design. [credits] Let’s say we’re doing an experiment where we see how different diets affect how much mice eat and how much weight they gain. One of the measurements we need to make is how much the mice weigh. But maybe each time we handle the mice we scare them so much that they don’t want to eat. We might think that the diet is working, but instead, it’s our measurement that is making a difference. What if we were trying to run a behavioral experiment where the outcome of interest is how children play with each other. If we put an adult in the room to record their behavior, we might frighten the children or change what they are doing in such a way that we see a difference, but it has less to do with our intervention than the measurement itself. These problems can be overcome! Think of the mouse study. We could choose to take measurements as rarely as possible. For instance, if we chose to measure weight only at the beginning and the end of the study, then our handling of the mice would be kept to a minimum, and so would its unwanted effects. In the same way, we could try to automate feeding and measure how much food the mice are actually consuming in such a way as to keep them unaware of our presence. If we’re feeling especially sophisticated, we could rig up scales in their cages so that weight could be measured without the mice even knowing about it. In our human study, the kids could be watched through a one-way mirror, so that they’re unaware of observers. Maybe that kind of environment is still too different to get real-world results. In that case, maybe we could set up monitoring methods in a playground where hidden cameras and microphones were used to monitor how kids played without their knowing it at all. As with many other forms of bias, thinking about them is not only the first step, it’s one of the most important ones. Acknowledging how we might be changing the results of our experiments is the only way to create ways to prevent us from doing so. Some of the ways that observer bias can affect experiments can’t be fixed. That’s ok. Considering and reporting these instances though will help other researchers to improve upon your work or at least be aware of our faults when they try and build upon it later. On the other hand, much observer bias can be overcome. Those methods are likely more expensive. They may be more time consuming. And they may be more difficult. But they’re important, and we should always consider them.