3 ms·

The problem, from where I sit, is that certain types of experiments have a built error detection or safety rails and psychology doesn't have this. For instance

The problem, from where I sit, is that certain types of experiments have a built error detection or safety rails and psychology doesn't have this. For instance if a physics experiment has a result which appears to violate conservation of momentum or has information being sent faster than the speed of life, then it is likely that there is a flaw in the experiment. Unfortunately many of the social sciences don't have these easily detected error conditions. Thus, it becomes much harder to determine is an experiment is flawed or not. Making this situation even worse social sciences are very hard to control for and replicate because you can't just order "one standard human" for the experiment. Two experiments may not replicate each other due to no fault of the experiment. What we need is a model of human psychology that provides clear boundary conditions.

fabatka 8y ago
But there are methods to deal with more or less all of these problems, that is why social scientists learn research design/methodology separately so much. In my opinion being able to replicate results should still be required to accept a theory. My problem with this case is that things that I consider basics for the scientific method (replication or even a control group) are obviously missing.
- EthanHeilman 8y ago
  To the people downvoting my comment I'd be genuinely interested in hearing why people see my comment as damaging to the discussion. >But there are methods to deal with more or less all of these problems There are methods to deal with designing experiments poorly, but even the best experiment could have some unknown unknown issue that skews the results. Physics has an inherent advantage that experimental mistakes often lead to results which contradict theory to such an extent that the experimenter can be fairly sure the experiment is in error. > In my opinion being able to replicate results should still be required to accept a theory. Strongly agree, however replication often becomes harder to achieve as the complexity of the system you are measuring increases. Photons are simple in ways that humans just aren't. For example various mental disorders only appear in certain cultures, the culture of a photon isn't a problem a physics experimenter needs to worry about. With studies on humans there are so many more places for failure to occur both in the original study and in the replication. Training matters, but the task is significantly more difficult.

fabatka 8y ago

But there are methods to deal with more or less all of these problems, that is why social scientists learn research design/methodology separately so much. In my opinion being able to replicate results should still be required to accept a theory. My problem with this case is that things that I consider basics for the scientific method (replication or even a control group) are obviously missing.

EthanHeilman 8y ago
To the people downvoting my comment I'd be genuinely interested in hearing why people see my comment as damaging to the discussion. >But there are methods to deal with more or less all of these problems There are methods to deal with designing experiments poorly, but even the best experiment could have some unknown unknown issue that skews the results. Physics has an inherent advantage that experimental mistakes often lead to results which contradict theory to such an extent that the experimenter can be fairly sure the experiment is in error. > In my opinion being able to replicate results should still be required to accept a theory. Strongly agree, however replication often becomes harder to achieve as the complexity of the system you are measuring increases. Photons are simple in ways that humans just aren't. For example various mental disorders only appear in certain cultures, the culture of a photon isn't a problem a physics experimenter needs to worry about. With studies on humans there are so many more places for failure to occur both in the original study and in the replication. Training matters, but the task is significantly more difficult.