Fudge Sunday - Twitter Travel Algorithmic Odditiesby Jay Cuthrell
A Tribe Called Quest - Scenario (1992)
This past week was a vacation where I had a chance to catch up on reading and test a what if scenario. While the reading was informative the what if scenario only served to compound my assumptions and generate new questions.
While reading, the topic of Twitter (drama?) kept creeping into my reading because of a variety of topics. So, this week I tested a what if scenario for basic web browsing to see if websites like Twitter might render different recommendations based on where I was spatially in a fixed temporal window.
This issue is thick with opinions (mine) but I’m also including several links to primary sources along the way. So, let’s get started.
Tell me, yo – what’s the scenario?
Perhaps you’ve had a super tracker experience. A simple example of a common super tracker experience is one where you search for a product online and find similar or directly related advertising for the product you searched for that seems to follows you around well after the search.
Perhaps you recall my blog post from 2017 about low leveling my Twitter account(s) after a decade. Now, looking back 5 years later, I’m still not sure how I really use Twitter or gain utility.
I don’t actively participate in Twitter in 2022. However, if you had told me the Eternal September Bots / Trolls and NFT reply spam would be this bad, I’d have assumed you were talking about 2000s era email and never looked back again or doubled down on Block Party (I used it successfully in 2021).
An examination of my Twitter archives
Anti-harassment and safety tools that block trolls and other abuse on Twitter via filters, block lists, and more. Take back control of your online experience.
You might also recall that I’ve worked in telecom and online everything since the late 1990s. So, I have some opinions as would many others if not most others.
What I’m not, is a candidate for extolling the virtues of strong opinions loosely held aka SOLH pigeonhole or similar neologisms or quips that permeate echo chamber enclaves – indeed the very pigeonholes themselves. So, with that in mind I’m comfortable saying that this newsletter issue and blog posts will (ideally!) not age well over time.
The point I hope to make is that things around us are in motion and in the narrowly defined scenario that I share below there are deep barrels of widely ignored things and plump juicy assumptions. Do not take my word(s) for it.
Before we get too far into the weeds, check out these links as a deeper dive before we get to the scenario:
IP targeting provides an unprecedented level of specificity for marketing professionals. Learn how to incorporate IP targeting into your marketing strategy.
Emerging data suggest that though people altered their habits during the first month of America’s response to the pandemic, that cooperation has since leveled off and — eventually — decreased.
Skyhook’s opt out options.
Learn about SafeGraph’s compliance with the California Consumer Privacy Act of 2018 (“CCPA”).
Information on Opting out on Mobile Devices: This page is provided to help more easily find consumer choices on various devices. It is provided for educational purposes only.
Techlore is spreading spreading privacy and security to the masses. Home of Go Incognito, Surveillance Report, VPN reviews, video tutorials, software/hardware reviews, communities, and more; join us today!
John Oliver’s “data brokers” show explores many privacy & security concepts. Let’s dive into commentary regarding the episode.
Okay. Let’s get to it.
Never on the left, cuz my right’s my good ear 🎶
Back in January 2022, I wrote in Belatedly Beckoning Bespoke Beacons that my Twitter archives included this gem just over a decade ago:
“BREAKING NEWS: RADIOS WITHIN MOBILE DEVICES COMMUNICATE WITH THINGS THAT HAVE A KNOWN LOCATION” – @qthrul 2011-04-22
Okay, let’s test that one in a very limited way.
- IP endpoints in a rural and metropolitan area (aka traveling)
- Cache cleared browser, no authentication, private/incognito mode
- Browse directly to https://twitter.com/explore
First, we test from an IP address that’s close to a metropolitan area.
Twitter recommended following these accounts with 2M Followers, 32M Followers, and 25.1M Followers respectively. So, I have no further commentary other than while these are massive following numbers none of these suggestions are in the Top 50 followed accounts on Twitter.
For perspective, to be in the Top 50 would require +37.2M Followers as of this issue of the newsletter.
Second, we test from an IP address that’s +100 miles from a city center.
Twitter recommended following these accounts with 2.6M Followers, 2.7M Followers, and 2.7M Followers respectively. Again, I have no further commentary other than I was nowhere near a city in Ohio or the state of Ohio at the time of this screenshot.
Again, for perspective, to be in the Top 50 would require +37.2M Followers as of this issue of the newsletter.
Causin’ rambunction throughout the sphere 🎶
Okay. Now that I went through this scenario – I wish I had done tests across different mobile devices, different OS devices, and generally fuzzing the browser environment variables as endpoints too. Some questions came to mind…
- Would different cost networks give different results by default?
- Would a new account profile give different spatial temporal results?
- Would an interaction between new accounts give different results?
- PROFIT!! (wait… no. wrong meme.)
But I digress.
So, it’s important to consider how a service you use is making money – not just from the advertising which is common – but as a function of the simulacrum for recommending what you should engage with now or next. Because, the alignment of incentives might not be anymore nefarious than a machine learning model that maps engagement based on the most benign basic banal assumptions when there is scant information or limited input for a so-called recommendation to present as now or next actions to take.
Our Promoted Ads are pay-for-performance or pay-for-impression delivered advertising that are priced through an auction.
Recommended Read instead of Repo?
This week I had hoped to link to https://github.com/twitter/the-algorithm but that’s not possible as of this issue since the empty repository has either been moved to private or deleted or it was an internal signal of artistic protest or none of the above – I have no insider information to share.
Instead, I’m embedding a tweet that is as lyrical as it was informative since it linked to a repository that was empty – and a great read from exactly a decade ago on the algorithms in practice at Twitter Trends circa May 2012.
watch this space
My most popular tweet about open source is for an empty repo. https://t.co/XyznhiMN2f
Tarleton Gillespie explores the controversy over Twitter Trends and the algorithmic ‘censorship’ of #occupywallstreet.
In other words, do not take my word(s) for it. Or, said another way, do not take the surfaced suggestion and suggested word(s) from a service for it either.
My recommendation is to look for primary sources that are less about opinions (like the opinions I’ve shared in this issue) and seek out multiple critiques that rely on widely vetted information. that is anchored in data-driven journalism which also seek to combine primary sources with repeatable research outcomes.
So, of course, this is easier said (1x) than done (10x)? And perhaps this means we are all living in the early stages of a period that can be described or summarized by Brandolini’s law.
Back in late 2021 I shared my thoughts on Twitter, Matter, and Data-Driven Journalism. Indeed, using Twitter Lists is still the only reasonable way I’ve found to consume Twitter in 2022.
Perhaps the future Twitter will alter the contract with users by leaving behind what I can only assume is gaming the feed for sheer outrage anchored engagement to manufacture advertising clicks and impressions. Again, perhaps.
Brandolini’s law, also known as the bullshit asymmetry principle, is an internet adage that emphasizes the effort of debunking misinformation, in comparison to the relative ease of creating it in the first place. It states that “The amount of energy needed to refute bullshit is an order of magnitude larger than is needed to produce it.”
I am linking to my disclosure.
✍️ 🤓 Edit on Github 🐙 ✍️
Get Fudge Sunday each week