Identifying Political Bias in the Instagram Reels Algorithm

Research with @JulesLemee under the supervision of Joseph Vybihal completed as part of the COMP 396 at McGill University.

Scrapers

Avoiding Instagram's anti-bot detection and anti-scraping measures is the big roadblock in this project. As data became even more valuable thanks to the explosion of LLM, companies like Meta have put in place much stricter anti-scrpaing measures, making scraping an uphill battle.

1. Simulated User Scraper

Activate it by running mainsim.py. All our credentials for created accounts for webscraping are in input/db/users.csv and are accessed by row number in mainsim.py.

Run
cd \Users\Admin\Programs\PycharmProjects\comp396 (wherever your project root is)
to get to your project's root

Then run:
C:\Users\Admin\Programs\PycharmProjects\comp396\.venv\Scripts\python.exe -m src.maincol.py

You modify the scraper settings in the inputs of the scraper function in mainsim.py.
mainsim.py uses doomscroll.py

This will open a browser window and start scraping reels: it will perform actions on the reels based on the parameters you set in mainsim.py's scrape() call. Watch out, most actions will get you flagged as a bot and either disable features (you wont be able to like or follow) or ban you outright (if you leave the same comment many times, it bans you instantly)

Parameters	Editable?	Info
username, password	dont touch directly	credentials in creds[], change row number only
session	dont touch at all	session number from counter.txt file
watch_time_percentage	touchable	% of reel duration spent watching
liked	touchable	choose whether to like the reel or not
pos_comment_left	touchable	choose whether to leave a positive comment or not put "" for no comment write your comment as a string to comment it e.g. "Great video!" WARNING Instagram will ban you instantly.
followed	touchable	choose whether to follow or not
shared	touchable	choose whether to share the reel or not
saved	touchable	choose whether to save the reel or not
profile_visited	touchable	choose whether to visit the profile or not
neg_comment_left	touchable	choose whether to leave a negative comment or not put "" for no comment write your comment as a string to comment it e.g. "Bad video!" WARNING Instagram will ban you instantly.
clicked_not_interested	touchable	choose whether to click not interested or not
quit_after	touchable	number of reels to scrape this session
condition	touchable	CONDITIONAL BEHAVIOR: if parameter condition is 1, then do everything for every reel if parameter condition is 2, then only do everything if the uploader is in the dataset AND the political bias is the same
political_bias	touchable	if 'L' then only do things if the uploader is in the dataset AND the political bias is left if 'R' then only do things if the uploader is in the dataset AND the political bias is right if '' whatever

2. Political Account Collector

Activate it by running maincol.py. All our credentials for created accounts for webscraping are in input/db/users.csv, and are accessed by row number in maincol.py.

This is a database builder that collects political accounts. It opens the uploader's profile and screenshots it then saves it to an AWS S3 bucket to make it accessible via a URL. It passes that URL to Instagram's API to get the political bias of the account. That data is added to the database.

This is so that when we run the simulated user scraper (mainsim.py), we can check if the uploader is in the database and what their political bias is. This is how we would evaluate whether the algorithm is biased or not.

We are limited only by anti-bot and anti-scraping measures. If you can get over this hurdle, you're golden.