The new test turned out as I had expected — more views and engagement on photo art only with emphasis on color saturation and complex linework. What was most interesting was that Stage A did not really respond the way I had assumed, but Stage B definitely fed the content to the generalized feed.
| Content Id | Format | Line Complexity | Saturation Level |
|---|---|---|---|
| 1 | photo art | high | high |
| 2 | photo art | high | high |
| 3 | photo art | high | high |
Content in reverse order.
Stage A Results:
| Content | Views | Interactions |
|---|---|---|
| 1 | 2 | 0 |
| 2 | 2 | 0 |
| 3 | 4 | 0 |
Stage B Results:
| Content | Views | Interactions |
|---|---|---|
| 1 | 7 | 5 |
| 2 | 16 | 2 |
| 3 | 12 | 1 |
What’s notable about this test is that Stage A only recommends to my followers and the majority of these followers on that stage are people that I know. On the other hand, Stage B is characterized by a distribution of players I do not know. Of a subset of those players, I noted repeated interactions from players on certain themes based on how the content was categorized. That’s useful, since I can use the same material, same theme but reach different cohorts.
Stage A is consistent with a limited and fixed network of sharing, less useful for observing the black box stage. Given the pattern, Stage B stands out as more dynamic for further exploration. The next step is adding more saturation and doing a test where I draw between categories and post-process it to convert from photo art to digital art.
Content in reverse order.
Well, this test turned out even more pronounced with the latest data! The results for Stage B are substantially different:
Stage A Results:
| Content | Views | Interactions |
|---|---|---|
| 4 | 4 | 0 |
| 5 | 5 | 0 |
| 6 | 4 | 0 |
Stage B Results:
| Content | Views | Interactions |
|---|---|---|
| 4 | 39 | 5 |
| 5 | 17 | 2 |
| 6 | 9 | 3 |
Stage A’s views and interactions were pretty consistent with the last test but Stage B has enough to give us a clue as to how they categorize. I even gained a follower from this last Stage B iteration. Could it be that the more abstract it appears to be, the more the content is fed into additional generalized feeds? I guess we’ll have to do more tests over time to find out! As for now, this exploration goes to the background, while I investigate another game.
This new game is a different type of analysis, not with social media and graphical content but only code on Github. I call this new game the “Strange Observer Effect”.