I am reading a research paper in which the authors designed three bots and wanted to determine what bot was most human-like (link is at the bottom). To do this they had subjects play against each bot and answer questions about each bots performance. The questions where answered on a scale of -2 to 2 where -2 was strongly disagree and 2 was strongly agree.
Once all the subjects answered the questions the authors averaged all of the responses and plotted them on a graph. The graph shows the mean agreement and has error bars that show 95% confidence intervals around the mean.
The authors then perform a 2-way analysis of variance on the bot type and question.
The results are as follows:
- There is a difference between bots
- The effect of the question is not significant.
- There is a significant interaction between bot type and question.
I have a weak background in statistics and I was hoping that someone could explain what these results mean in simple terms.
http://ro.ecu.edu.au/cgi/viewcontent.cgi?article=1732&context=ecuworks
