---
tags: Book
---

# Human Evaluation Spreadsheet Instructions
Thank you for agreeing to support Gitcoin's anti-Sybil team!

These notes will be included in the Google sheets that contain the usernames you are supposed to evaluate for Sybil activity. They are provided here as a courtesy and backup.

A [video explaining the process](https://share.getcloudapp.com/4gun4mqz) is also available.

Each Tab is for a different reviewer to use. If a tab has "HEA_test #" it is unassigned. Please choose an unassigned tab and rename it replacing the "test #" with your name and then follow the instructions below.

Evaluators are expected to evaluate at least 15 profiles; you are welcome to do more, if you like.

We are looking for users and accounts that may have been set up to game the Gitcoin funding round and gain more funds than they should have.

On your tab please fill out columns G-J. In order, they ask the following questions:

G) Is this a Sybil account?
a. Yes
b. No
This question is asking, in your opinion as a human, whether the account you're looking at is part of a Sybil attack. If you don't know, then just guess and make sure that you mark the answer as being low confidence.

H) What is your certainty about the answer you gave in (G)?
a. low = Not certain at all.
b. so-so = Not completely certain, but would bet on it.
c. high = I AM certain about my answer
This question is asking whether you are sure about your answer to the previous question. Uncertainty is OK! Sybil attacks are difficult to identify.

J) A notes column to explain your decision very briefly

**All four columns are required.** Please note that columns (H) and (I) relate to each other.

For example:

If you believe that the account you're evaluating is a Sybil account, AND you are certain in that belief the columns would look like this:
|account| is_sybil | certain | sybil_score | notes |
| --- | --- | --- | --- | --- |
| some_account | 1 | 1 | 5 | just looks funny |

On the other hand, if you look at the account and think it's a Sybil account but are not certain then the columns might look like this:

|account| is_sybil | certain | sybil_score | notes |
| --- | --- | --- | --- | --- |
| some_account | 1 | 0 | 3 | just looks funny |

Alternately, if you look at the account and you're certain that it is not a Sybil account, you might fill out the columns like so:

|account| is_sybil | certain | sybil_score | notes |
| --- | --- | --- | --- | --- |
| some_account | 0 | 1 | 1 | looks pretty human |

Regardless, one would expect that a 1 in the `certain` column would result in a 1 or a 5 in the `sybil_score` column, as those are the strongest scores on either end of the Sybil/Not-Sybil spectrum