# Political TikTok Project

##### To Discuss

+ Frequency of scraping
+ TikTok users (every three days)
+ Political Hashtags (every day)
+ Downloading TikToks (1.3 TB)
+ Will be integrated into scraping routine
+ Labeling of TikToks
+ left - right
+ socialist/communist - libertarian
+ RA Recruiting for labeling?
+ Generate new list of hashtags
+ "Shadowbanning"
+ Problem: people keep create new accounts and delete old ones
+ Trump Tulsa Rally
+ Not a lot of users talk about it in TikTok description
+ Potentially need to extract text from pictures/audio
+ Text data
+ Comments, Descriptions, Bios
+ Extract text from pictures
+ Transcribe audio text

##### TODO

###### Technical
+ What is the life cycle of (political) TikToks?
+ How long do TikToks get comments?
+ Test: check the comment cycle
+ How long do trends on TikToks last?
+ Test: scrape trending TikToks once a day from different location/region
+ How stable are hashtag query results?
+ Across time and space
+ Test: scrape once a day a list of 100 hashtags from different locations/regions

Table of Content

* [Objectives](https://hackmd.io/8UsEWgzoRzqYxPEBPA9mOA?both#Objectives)
* [Theoretical Framework](https://hackmd.io/8UsEWgzoRzqYxPEBPA9mOA?both#Theoretical-Framework)
* [Sampling](https://hackmd.io/8UsEWgzoRzqYxPEBPA9mOA?both#Sampling)
* [Ideas](https://hackmd.io/8UsEWgzoRzqYxPEBPA9mOA?both#Ideas)
* [Tools](https://hackmd.io/8UsEWgzoRzqYxPEBPA9mOA?both#Tools)
* [Further Reading & Lit](https://hackmd.io/8UsEWgzoRzqYxPEBPA9mOA?both#Further-Reading-amp-Lit)

## Objectives

What are the objectives of this project? Research question?

+ Descriptive Analysis of Political TikTok
+ aggregate popularity of left v right;
+ trends in popularity over time;
+ topics;
+ (relative size of pol. TikTok to rest of TikTok?)
+ (comparison of TikTok to other platforms)

## Theoretical Framework

What kind of theoretical framework do we use?

+ Supply & Demand?
+ Supply-side: Why should user create political TikToks?
+ TikTok users turning political or political "figures" turning to TikTok?
+ What is the role of traditional political actors?
+ Access to a specific population (that is hard to reach)
+ Who are "political" users?
+ Politics as an infinite and often renewed source of inspiration, which already has a community. Maybe easier as other topic to be trending
+ Content can be created easily from home
+ Short succinct videos
+ Blurry line between consumer and creator (everyone can be creator)
+ Demand-side:
+ Entertainment
+ Simplification of political messages to keep it entertaining?
+ Identity reinforcement ("Those liberals/conservatives")
+ Confrontation between two sides
+ Does it work the same way in party systems with more than two parties?
+ Politics

## Sampling

How do we get a decent sample of political TikTok?

### Current Approach

1. Explore political TikTok (ask people from community)
2. Create a list of around ~100 accounts and manually label by ideology
3. Create a list of political hashtags (~150 hashtags)
4. Sample TikToks that mentioned political hashtags (~33k TikToks)
5. Create a list of accounts that mentioned political hashtags (~1000 TikTokers)
6. Label all accounts by ideology and content
+ In total: +965 accounts identified as either right or left leaning
7. Get all TikToks from +965 political accounts
+ In total: +228k TikToks

- Create a spreadsheet with political TikTokers ([Master List](https://docs.google.com/spreadsheets/d/13kg6HELkzHgykmr2_8wkI2AI-PbMIq2yfx2UHttmZPQ/edit?usp=sharing))
+ Ask people who consume (political) TikTok for their input
+ 100 to 1000 individuals would be great
+ Validation
+ Query political hashtags and see how many more users appear than we already have in our "master list"

#### Labeling

Check the 10 most recent TikToks and see whether they are political or not. If yes, check political affiliation (left, right).

### Old Approach

Our old is approach is just "experimental" and was used for developing the `tiktokr` API.

1. Scrape all available TikToks for a small list of political hashtags: #votebiden, #votetrump, #trump2020, #bernie2020, #biden2020, #maga, #trump
2. Scrape all TikToks from each account that mentioned any of the selected hashtags in at least 20 TikToks
3. Update the list of hashtags by handpicking the most common political hashtags, repeat as necessary.

### Suggested Accounts

One possible avenue to explore is looking at suggested accounts for specific users. It's unclear though whether this differs by region and/or personal account.

### Duets

> What even are duets? TikTok Duets basically allow users to create content featuring an initial video, with both videos appearing side by side on screen in a square format. This means that people can effectively reply to video content with their own videos – which then can be shared publicly.

Duets could be used for sampling, network visualization and/or community detection.

### Limits & Issues

+ **Using Accounts suggestion for sampling:**
+ After in depth search, it appears that account suggestion are only marginally affected by the current visited user. It is mostly affected by popularity and is not suited to sample political users
+ **Using Hashtags for sampling:**
+ Query through hashtags only provides 2k hits, which are not drawn randomly or based on the most recent post date but rather **some mix of recency and popularity** of TikToks.
+ Hashtags are not just an "in-group community marker" but are also frequently used to attract people from the other political side.
+ **User sampling:**
+ Query through users often only provides 2k hits but limit is unclear. Sample seems to be from most recent to oldest.
+ **Text data:** Text data from comments is available but often full of spam-like copy-paste text, very short sentences, memes and slogans.
+ **Robustness of Scraper** It's possible that TikTok changes their infrasctructure again so we need to account for that (and move fast)

## Ideas

Some additional ideas, research questions?

+ Politicization of TikTok during BLM protests (or other prior events):
+ Question: Can we observe a politicization over time?
+ Look at individual TikToker trajectories
+ Automated Transcription of TikTok Videos
+ Possible but a lot of TikToks only use music
+ Extract text from Video/Pictures
+ Topic modeling on text data
+ Possible but text often gibberish
+ Get mainstream media accounts
+ e.g. https://www.tiktok.com/@washingtonpost
+ Stream TikTok data all the way through 2020 Presidential Election
+ Other Platforms to investigate
+ Instagram
+ Twitch (political Streamer community)
+ Other countries:
+ UK, Germany, France
+ What music do liberals vs. conservative use?

## Tools

+ [`tiktokr`](https://github.com/benjaminguinaudeau/tiktokr)
+ [Social Blade (for checking TikTok Stats)](https://socialblade.com/)
+ [HASHTAG GENERATOR FOR TIKTOK: Enter your hashtag and we tell you the best related one.](https://tiktokhashtags.com/)

## Further Reading & Lit

### Academic

+ [Munger, Kevin, and Joseph Phillips. "A supply and demand framework for Youtube Politics." Preprint on OSF (2019).
](https://osf.io/73jys/)
+ [Serrano, Juan Carlos Medina, Orestis Papakyriakopoulos, and Simon Hegelich. "Dancing to the Partisan Beat: A First Analysis of Political Communication on TikTok." arXiv preprint arXiv:2004.05478 (2020).](https://arxiv.org/pdf/2004.05478.pdf)

### Articles

+ [Vice News: Young Aspiring Boogaloo Bois Are Dancing With Guns on TikTok (June 16th 2020)](https://www.vice.com/en_us/article/xg8yjw/young-aspiring-boogaloo-bois-are-dancing-with-guns-on-tiktok)
+ [CNN: TikTok serves as hub for #blacklivesmatter activism (June 4th 2020)](https://edition.cnn.com/2020/06/04/politics/tik-tok-black-lives-matter/index.html)
+ [TheConversation: How TikTok got political (2nd June 2020)](https://theconversation.com/how-tiktok-got-political-139629)
+ [FT: TikTok becomes political platform ahead of US election (2nd June 2020)](https://www.ft.com/content/c4c09793-993e-4ffd-9e46-2c609f98b79d)
+ [NYT: The Political Pundits of TikTok (27th February 2020)](https://www.nytimes.com/2020/02/27/style/tiktok-politics-bernie-trump.html)
+ [Vox: TikTok never wanted to be political. Too late. (January 22nd 2020)](https://www.vox.com/the-goods/2020/1/22/21069469/tiktok-memes-funny-ww3-politics-impeachment-fires)
+ [Morning Consult: New Progressive Campaigns Are Trying TikTok Despite Establishment Reluctance (16th January 2020)](https://morningconsult.com/2020/01/16/new-progressive-campaigns-are-trying-tiktok-despite-establishment-reluctance/)
+ [WSJ: TikTok Wants to Stay Politics-Free. That Could Be Tough in 2020. (January 5th 2020)](https://www.wsj.com/articles/tiktok-wants-to-stay-politics-free-that-could-be-tough-in-2020-11578225601)
+ [Mic.com: Is TikTok a looming political disaster? (19th November 2019)](https://www.mic.com/p/is-tiktok-a-looming-political-disaster-19354545)

+ List of political Hashtags
`"2a", "2ndamendment", "acab", "amerikkka", "anarchy", "aoc",
"barbz4bernie", "bernie_", "bernie2020", "berniesanders",
"biden", "biden2020", "blackconservative",
"blacklivesmatter", "blackrepublican", "blexit",
"blm", "climatestrike", "communism", "communist",
"conservativegirls", "conservativehypehouse", "conservativerepublican",
"conservatives", "creepyjoebiden",
"democraticsocialism", "democrats", "donaldjtrump", "donaldtrump",
"donaldtrump2020", "dumptrump", "eattherich", "fakenews",
"feelthebern", "feminism", "feminist", "georgefloyd", "guncontrol",
"impeach", "impeachment", "impeachtrump",
"joebiden", "justiceforgeorgefloyd", "kag2020",
"kavanaugh", "keepamericagreat", "leftist", "leftists", "lgbtrights",
"liberal", "liberalhypehouse", "liberalism", "liberals", "liberalsaredumb",
"libertarian", "libsoc", "lockhimup", "maga", "maga2020", "magachallenge",
"notmeus", "obama", "openyoureyes_", "political", "politicalhumor",
"politicallyincorrect", "potus", "presidenttrump", "pro2a", "prochoice",
"progressive", "prolife", "protest",
"republican", "republicanhypehouse",
"republicans", "soc", "socialism",
"socialist", "tedcruz", "thanksobama", "tpusa",
"transrights", "triggeredliberal", "trump", "trump10to20",
"trump2020", "trump2020kag", "trumppence", "trumpsupporters",
"trumptards", "trumpteam", "trumptrain", "trumptrain2020",
"trumpvoters", "twogenders", "usgovernment", "uspolitics", "vote",
"votebiden", "voteblue", "voteblue2020", "votebluenomatterwho",
"votetrump", "warren2020", "whiteprivilege", "woke",
"wokeposting", "women4trump", "womenfortrump", "ww3",
"yang", "yang2020", "yanggang"`