Skip to content

Commit

Permalink
Conspiracy Theories
Browse files Browse the repository at this point in the history
  • Loading branch information
@BurntWaffleCake
BurntWaffleCake committed Oct 22, 2023
1 parent dcd3e7e commit 6044f1d
Show file tree
Hide file tree
Showing 4 changed files with 191 additions and 56 deletions.
16 changes: 8 additions & 8 deletions CS-III/Projects/AdvancedSorts.html
View file
Original file line number Diff line number Diff line change
Expand Up @@ -14,19 +14,19 @@

<style>
.image-holder {
max-width: 100%;
max-width: 100%;
}

.image-holder img {
max-width: 100%;
height: auto;
background-color: white;
max-width: 100%;
height: auto;
background-color: white;
}

.article-nav-header {
font-size: larger;
font-size: larger;
}
</style>
</style>

<body>
<object id="global-nav-object" data="../../articleTemplateObjects/articleGlobalNavObject.html"></object>
Expand Down
226 changes: 178 additions & 48 deletions CS-III/Projects/ConspiracyTheory.html
View file
Original file line number Diff line number Diff line change
Expand Up @@ -3,59 +3,189 @@
<html>

<head>
<meta charset="utf-8">
<meta http-equiv="X-UA-Compatible" content="IE=edge">
<title></title>
<meta name="description" content="">
<meta name="viewport" content="width=device-width, initial-scale=1">
<link rel="stylesheet" href="../../styles/articlestylesheet.css">
<link rel="stylesheet" href="https://fonts.google.com/specimen/Lora">
<meta charset="utf-8">
<meta http-equiv="X-UA-Compatible" content="IE=edge">
<title></title>
<meta name="description" content="">
<meta name="viewport" content="width=device-width, initial-scale=1">
<link rel="stylesheet" href="../../styles/articlestylesheet.css">
<link rel="stylesheet" href="https://fonts.google.com/specimen/Lora">
</head>

<style>
.image-holder {
max-width: 100%;
}
.image-holder img {
max-width: 100%;
height: auto;
background-color: white;
}
.article-nav-header {
font-size: larger;
}
</style>
.image-holder {
max-width: 100%;
}

.image-holder img {
max-width: 100%;
height: auto;
background-color: white;
}

.article-nav-header {
font-size: larger;
}
</style>

<body>
<object id="global-nav-object" data="../../articleTemplateObjects/articleGlobalNavObject.html"></object>

<div id="article-body">
<object data="../../articleTemplateObjects/csArticleNavObject.html"></object>

<div class="vertical-divider"></div>

<article id="article-contents">
<h1>Conspiracy Theories</h1>
<h2>The Basics of Recommendation Algorithms</h2>
<h2>The Feedback Loop</h2>
<h2>Conspiracy Theories & Algorithmic Amplification</h2>
<h2>Platform Incentives</h2>
<h2>Potential for Exploitation</h2>
</article>

<div class="vertical-divider"></div>

<nav id="article-bookmarks">
<a href="#Abstract" class="article-nav-link article-nav-header">Conspiracy Theories</a>
<a href="#Abstract" class="article-nav-link article-nav-header">The Basics of Recommendation Algorithms</a>
<a href="#Abstract" class="article-nav-link article-nav-header">The Feedback Loop</a>
<a href="#Abstract" class="article-nav-link article-nav-header">Conspiracy Theories & Algorithmic Amplification</a>
<a href="#Abstract" class="article-nav-link article-nav-header">Platform Incentives</a>
<a href="#Abstract" class="article-nav-link article-nav-header">Potential for Exploitation</a>
</nav>
</div>
<object id="global-nav-object" data="../../articleTemplateObjects/articleGlobalNavObject.html"></object>

<div id="article-body">
<object data="../../articleTemplateObjects/csArticleNavObject.html"></object>

<div class="vertical-divider"></div>

<article id="article-contents">
<h1>Conspiracy Theories</h1>
<h2 id="01_00">The Basics of Recommendation Algorithms</h2>
<p>
Recommendation algorithms are commonly used in online applications such as social media, ads, and marketing
sites. These algorithms are designed to find the best fitting data sets and match them to people most likely to
relate and invest in them. Companies use the algorithms for applications such as recommending relevant products
in advertisements or finding related topics when searching for certain ideas (as Google does when you search for
things). Who main recommendation paradigms exist: collaborative and content. Each conform to different datasets
and have their own methods of categorizing information and compiling such information into results.
</p>

<h3 id="01_01">Collaborative Filtering</h3>
<p>
Collaborative filtering uses past interactions between users and items to produce new recommendations and
relevant data. Past interactions are stored in a "user-item interactions matrix" which holds different metrics
regarding user's interactions with the item (rating, time spent on data, click interactions, etc). These
matrices are then used to estimate similar items and/or users to group data points into clusters of similar
information.
<br />
<br />

Collaborative filtering is divided into two subcategories: memory based and model based. Memory based approaches
work directly with the matrix's raw values and search for similar data (find similar users by comparing matrix
values and recommend items these neighbors find interest in). Model based approaches use an underlying
generative model that tries to predict new related data points using that model.
<br />
<br />

Some benefits that collaborative filtering has is that it requires no starting data to implement and instead
reads information that users generate as they interact with the product. This means that more interactions and
time result in better recommendations due to the dataset getting larger and more interactions are held. However,
this does mean that collaborative filtering is weak during the early stages of the product where no data about
its users are present meaning no recommendations can be made to new users. This problem is also present in users
with very little interactions which will result in inaccurate predictions by the algorithm. To alleviate this
effect, random or new data points can be chosen and given to random users which is called the random strategy.
Another strategy used is the maximum expectation strategy where generally popular data points or completely new
data points are given to the most active users. Along with this, the exploration strategy can also be used where
sets of various items with similar structures are given to new users or new items are given to similar user.
</p>

<h3 id="01_02">Content Filtering</h3>
<p>
Content filtering uses additional available data such as age, sex, occupation, and other initial data to form
recommendations. Content filtering build models off of these available datasets to recommend data points using
observed user-item interactions. For example, if a person generally watches action oriented films and movies
more than other genres, the model will recommend more actions films to that person.
<br />
The initial issues present in collaborative filtering is less pronounced in content filtering as new data points
can be recommended to new users using already available data on them. The problem of completely new users or
ones with previously unrecorded interests persists; however, these drawbacks are solved as more and more data is
added into the system.
</p>

<h2 id="02_00">The Feedback Loop</h2>
<h3 id="02_01">What Is It?</h3>
<p>A feedback loop is a part of a system where outputs of certain operations are used as inputs of the same
operation in the future. Recommendation algorithms use feedback loops to iteratively generate new data points
for the system to use as stepping stones to develop more accurate and relevant topics users may find interest
in.</p>

<h3 id="02_02">Feedback Loops in Recommendation Algorithms</h3>
<p>
Most recommendation algorithms use some sort of built dataset to recommend new information to users. These
datasets are created using the already available data about the user or are generated and stored as the user
interacts with the product. The growth of these datasets are what cause feedback loops to form where new data is
closely related to older data forming biases and causing the algorithm to focus more on those similar topics.
This causes an effect referred to as the "filter bubble" or "echo chamber" where very similar topics are
reiterated to the user in increasing intensity causing the user interacts with the same topics and forming a
feedback loop.
</p>

<h3 id="02_03">So What?</h3>
<p>
As companies try to get their users to engage more with their product, many users will find themselves getting
recommended biased or "generic" content that conforms to their interests. This may lead to tunnel visioning or
false belief in topics that fail to branch out into other relevant or new topics which may lead to
misinformation, misrepresentation, among many other problems.
</p>

<h2 id="03_00">Conspiracy Theories & Algorithmic Amplification</h2>
<h3 id="03_01">Conspiracy Theory Catastrophe</h3>
<p>As companies look for ways to increase user engagement and metrics such as look time, shares,
clicks, etc. a general trend that has emerged is that people tend to engage much more with sensationalistic
content such as conspiracy theories or conflict / internet drama. The fluctuating, diverse, and magnitudal cast
of user reactions to these types of content had resulted in an increased priority for such content to be shared
with newer and engaging people. As discussed in the paragraphs above talking about recommendation algorithms and
feedback loops, this increased level of engage builds up at rates much faster than regular articles and content
which in turn recommends these topics to newer and more individuals increasing engagement even more.
<br />
<br />
What this means is that topics regarding very polarized opinions or outlandish and outright wrong opinions can
be spread extremely quickly through increased user interactions resulting in mass misinformation and
misunderstandings. This is further emphasized by the general laziness of the average internet surfer, many who
rely on single, unreliable sources, and who believe what others say without much thought.
</p>
<h2 id="04_00">Platform Incentives</h2>
<h3 id="04_01">The Basis of the Problem</h3>
<p>
Many of the companies and applications where this problem exists use user interactions as their main source of
relevancy and income which only incentivize the promotion of these types of conversations and topics even more.
This type of behavior especially plagues information outlets such as news and article websites where titles are
often misleading, hyperbole is generously sprinkled throughout, and information is just outright false. The fact
that these types of articles are what people find most interesting and the strategies used to generate visits is
what fuels the engine for conspiracy theories and other types of misleading content to flourish.
</p>
<h2 id="05_00">Potential for Exploitation</h2>
<h3 id="05_01">Escalation</h3>
<p>
In many cases, the misinformation spread throughout the internet through conspiracy theories and other forms of
sensationalistic articles are generally ridiculed, made fun of, and passed on as outlandish ideas that nobody
would believe in. However, the implications and powerful effects that these types of topics can have can easily
be used to exploit the human mind. Objectives such as propaganda, targeted attacks, and many other forms of
dangerous and destructive behavior can lead to devastating consequences.
</p>
<h3 id="05_02">What Can We Do?</h3>
<p>
Most of the problems present in conspiracy theories and misleading information hubs can be alleviated through
thorough researching and exploration. As a matter of fact, many of these theories only require a line of
reasoning or common sense to solve. Making sure to check multiple credible sources of information, keep an open
mind, and carefully work through a concrete line of reasoning can easily knock out the most obvious falsities
and result in a deeper level of perception towards misleading information.
</p>

</article>

<div class="vertical-divider"></div>

<nav id="article-bookmarks">
<a href="#01_00" class="article-nav-link article-nav-header">The Basics of Recommendation Algorithms</a>
<a href="#01_01" class="article-nav-link">Collaborative Filtering</a>
<a href="#01_02" class="article-nav-link">Content Filtering</a>

<a href="#02_00" class="article-nav-link article-nav-header">The Feedback Loop</a>
<a href="#02_01" class="article-nav-link">What Is It?</a>
<a href="#02_02" class="article-nav-link">Feedback Loops in Recommendation Algorithms</a>
<a href="#02_03" class="article-nav-link">So What?</a>

<a href="#03_00" class="article-nav-link article-nav-header">Conspiracy Theories & Algorithmic
Amplification</a>
<a href="#03_01" class="article-nav-link">Conspiracy Theory Catastrophe</a>

<a href="#04_00" class="article-nav-link article-nav-header">Platform Incentives</a>
<a href="#04_01" class="article-nav-link">The Basis of the Problem</a>

<a href="#05_00" class="article-nav-link article-nav-header">Potential for Exploitation</a>
<a href="#05_01" class="article-nav-link">Escalation</a>
<a href="#05_02" class="article-nav-link">What Can We Do?</a>
</nav>
</div>
</body>

</html>
1 change: 1 addition & 0 deletions articleTemplateObjects/csArticleNavObject.html
View file
Original file line number Diff line number Diff line change
Expand Up @@ -85,5 +85,6 @@
<a href="../CS-II/Essays/AISentience.html" target="_top" class="article-nav-link">AI Sentience</a>
<a href="../CS-II/Essays/Decision Making.html" target="_top" class="article-nav-link">Decision Making</a>
<a href="../CS-II/Essays/FingerPrints.html" target="_top" class="article-nav-link">Fingerprints</a>
<a href="../CS-III/Projects/ConspiracyTheory.html" target="_top" class="article-nav-link">Conspiracy Theories</a>
</nav>
</body>
4 changes: 4 additions & 0 deletions styles/articlestylesheettemp.css → deprecated/articlestylesheettemp.css
View file
Original file line number Diff line number Diff line change
Expand Up @@ -177,6 +177,10 @@ body {
line-height: 1em;
}

#article-contents h2 {
text-decoration: underline;
}

#article-contents h2, h3, p {
margin-top: 1rem;
margin-bottom: 1rem;
Expand Down

0 comments on commit 6044f1d

Please sign in to comment.