Are computers bias

Computers have no emotions so how can they be bias. well they seem to prefer different colours in the same position. e.g.

The same position occurs with different colours yet the evaluation is different. Here is another example.

what do you think is happening?

The positions are not identical. For example in your second case, white has played a move that reduces their flexibility a great deal compared to the second case.

Madratter7 and Kemp7, you are incorrect.  The pairs of positions shown by chesspuzzlerjunior are logically equivalent.  I don't know what chesspuzzlerjunior's computer would say, but I had a phase where I was winning a lot of games as black with the Sicilian, so I used to annoy my opponents by playing as white 1. c3, e5; 2. c4.  To play this position as white is logically the same as playing 1. e4, c5 with black.  The computer must have some algorithm that gives an automatic slight advantage to white, but it is not logical.

In the first two positions, it seems to favor the side without the g-pawn pushed, as a "minus" sign means in favor of Black. Why it has different numbers I don't know.

 The positions aren't equivalent, because some side of position 1, 2 and 3 wastes a move on not developing their pieces.

I've come across this before when playing positions with colors reversed then analyzing with an engine.

You have to let it think a long time before it will return the same eval.

I mean... it should be instant of course, so by "long time" I mean ~30 seconds.

As for the technical explanation for why it happens, I don't know, a good question for a engine forum.

 omg

Clearing hash table might help (?) My engine only took a few seconds before returning 0.00 for each of these

It favored black in both cases...

I think they switched over to all Komodo (?)

But website engines are dopey because they run in the browser and only think for a few seconds.

Thanks to preggo, james coleman, patzermike and chessically inclined. The others didn't seem to realise it was the same positions but with colours changed. I have checked it on a number of engines on other sites and they also seem to present different evaluations. I dont know what it is happenning but i am sure that somewhere out there there lies a logical explanation.

Interesting observation! The reason could be that an eval is based on the final position of a line of analysis, and Stockfish is actually giving you the evals for two different positions. Here are screenshots for Stockfish's analysis of your first pair of positions. The settings are the same for both: 1 line of analysis with a time limit of 30sec. Despite this, Stockfish stops analysing at different depths and the final positions (when reflected) are completely different. Indeed the two lines deviate very early: 3.e4 vs 3...Nc6. So the question becomes, why does Stockfish like these different moves in exactly the same situation?

Re-read the post. The same player (position-wise) has the move; the positions are fully identical.

Since 2018, stockfish programmers introduced contempt. 

This is the latest version of stockfish , whereas you can see contempt of "21" on top left coner.

Which means, in exactly equal position, stockfish in white will analyse as  +0.21 evaluation as white and -0.21 as black. (Making a difference of up to 0.42 evaluation from different views) 

Seemed wired! In fact the main reason was that in TCEC season 11, stockfish missed TCEC championship final , even though stockfish did not lose any game in premier division but drew a lot , (whereas Houdini and Komodo won too many games against weaker engines)

If you are really serious about those evaluation numbers, you can manually reset to "zero".

In earlier version of stockfish from in 2018 , up to a few months ago, the contempt was only "12"

If you use stockfish version in 2017, contempt was zero.

However, even though without contempt, engines evaluation will differ in depth, use of hash memory , use of table base etc. 

Alternatively, you can set up position and see how stockfish evaluate position in middle game and endgame, by using these evaluation tables.

If chess.com's auto analysis has contempt on... oh man...

Just another reason I don't use website engines.

Thanks LMPlayer Rocky and Rook sacrifice. But RedGirl why should eval be different. It is same position, same player to move except somebody came along with a paintbrush.