Automated Content Production

Hello and welcome to module 5 of the Automated Content Production and Algorithms course called – Algorithms in news distribution.

This module will cover;

1. How algorithms help in news distribution

2. How a change in algorithm can make or mar a news platform

3. Understanding algorithms based on their types of decisions

4. Algorithms as a gatekeeper of information

Also as a reminder, after watching the videos, please proceed to read the three documents that are added to this module.

After this, please ensure you take the quiz.

How algorithms help in news distribution

On May 25, 2020, George Floyd, a 46-year-old black man, was killed in Minneapolis, Minnesota, during an arrest for allegedly using a counterfeit bill. Derek Chauvin, a white police officer, knelt on Floyd's neck for almost nine minutes while Floyd was handcuffed and lying face down in the street, begging for his life and repeatedly saying "I can't breathe". A second and third officer further restrained Floyd while a fourth prevented bystanders from intervening. During the final three minutes, Floyd was motionless and had no pulse while Chauvin ignored onlookers' pleas to remove his knee, which he did not do until medics told him to.

By morning of the 26th Google searches for “George Floyd” were trending indicating a surge of interest as the public frantically searched for information and further details of what really transpired. Search engines and social media platforms played a key role in the dissemination of the news. But were those search engines providing accurate information? Were they providing diverse or balanced information? As the media rolled out headlines upon headlines and the news spread like wild fire, hoaxes, conspiracy theories and other falsehoods began spreading on Facebook and Twitter. Two very significant false hoods were posts suggesting that George Floyd was alive and that the Billionaire investor George Soros and donor to the Democratic party was funding the demonstrations against police brutality.

Now upon a proper investigation for the most likely source of news when a Google search is done on George Floyd for every minute for 24 hours, one would find a huge lack of diversity in the news sources. In the 24-hour period, the investigation would show a handful of the same news sources being responsible for the “Top Stories” menu on Google search results. One would discover that the top 6 of those news sources accounted for 85% of the links to stories that showed. The platforms include New York Times, Aljazeera, BBC, NBC News, CNN, Sky News.

Social media like twitter also played a key role as the ‘George Floyd went up the trends list. In news consumption, search engines and social media platforms play key roles as their algorithms determine the information the public see and engage with.

Winning and losing with algorithms

A change in Algorithm can shoot up the traffic on a news platform or it can reduce it drastically.The publishing platform Medium was launched in 2012 by Twitter co-founder Evan Williams. The platform uses an intelligent algorithm primarily based on the length of time users spend reading articles to recommend content to reader. The out turn of this is that users are going to find content that spans the reading time they are used to. This simple feature will keep the user coming back.

Google, Facebook and Twitter are highly responsible for referral traffic to websites. In 2017 Facebook decided to prioritize posts from friends and family in its newsfeed algorithm. This reduced referral traffic from Facebook by 15 percent, i.e., from 40 percent to 25 percent. This in turn dropped the traffic of several news platforms by several pegs.

Algorithms make it easier for people to find content and vice versa. The recommended stories you see at the end of every story on any news site are calculated via algorithm. These algorithms take into consideration the news content you click. The Times’s algorithm can display more than 160 different layout combinations, changing to accommodate different types of articles and images, and for emphasis. Small differences in labeling and typography allow the algorithm to use design to signal editorial importance

Algorithms and categories

As mentioned in a previous module, Algorithms can be categorized based on the types of decisions they make.

Prioritization, for example, ranks content, prioritizing one content over the other.

Association marks relationships between entities, such as articles or videos that share subject matter of features.

Filtering involves the inclusion or exclusion of certain information based on a set of criteria.

Classification ranks content based on the similarities of subject matter

Calculation, this can be viewed in terms of a ranking algorithm selecting articles by stat, for example by number of Up votes it has.

Algorithms as gate keepers

Algorithms play a key role in determining what is visible and what is not. When algorithms serve users contents based on certain decisions they make based on their consumption habits, that means that every other information that could have been useful to the user is excluded from their sights. Algorithms are responsible for shaping viewpoints on several societal issues even if these viewpoints are not true.

During the first waves of the George Floyd protests, there were coordinated attempts at misinforming the public circulating online; there were posts that read “George Floyd is alive, this an intentional scheme to start a race war”. Another read; “Get paid to be a Professional Anarchist!”. And algorithms would do what they do best, spread these information, hence helping to shape thoughts. Since many people searching for information online are not aware of the extent algorithms help drive agendas, they are unable to make the decision to look outside the information pocket provide by the algorithms.

Finally, due to these reasons, several traditional media platforms and non-governmental organizations who value truth and best practices take it upon themselves to constantly fact-check information, especially the viral ones. Today, many news platforms are aware of how algorithms influence their content and their readers, and as such they are developing their own curation algorithms, and proclaiming their freedom hence, shrinking their dependency on algorithms created by non-journalistic institutions.

In the next video I’ll get into Content Optimization and ways understand how your content performs. That’s it for now.

Hello and welcome to module 6 of the Automated Content Production and Algorithms course called – Content optimization and metrics.

This module will cover;

• Content optimization, an overview

• How publishers are augmenting the news dissemination

• Methods and methodologies to enhance the spread of news

• Tools to optimize the performance of content online

Also as a reminder, after watching the videos, please proceed to read the three documents that are added to this module.

They include; Know your algorithm; Social Media Platforms as News Publishers and Distributors; and Deepening Engagement for Lasting Impact.

After this, please ensure you take the quiz.

Content optimization, an overview

This is the practice of making sure that content is written in a way that it can reach the largest possible target audience. The process of optimizing includes ensuring that associated keywords are present, adding meta and title tags, and relevant links. This can also be the process in which a webpage and its content are optimized to become more attractive, useful and actionable to users.

Data processing algorithms are increasingly applied to audience data with an eye toward expanding the monetizable audience by predicting and optimizing the performance of content based on historical patterns. News publishers are constantly tracking their audience to study their consumption pattern and gain insight into reader behavior towards delivery of a more custom stories.

Audience analytics are a necessity in today’s newsrooms. Many newsrooms who take content optimization serious are following the data on their stories real-time. They have large flat-screen monitors projecting real-time metrics of the “trending” or “most read” stories. Insights garnered from these screens are used to create headlines, and content that they anticipate their audience want to see. A good example of this was a voter education and citizen engagement project I deployed in 2014 to 2105 for the general elections in Nigeria, except that this time, the process was in reserve.

Content Optimization- Project EDAV

The project was called EDAV – Election Data Analytics and Visualization. The goal of the project was simple; find key data and information that is relevant to the electorates’ needs and make it easily accessible to them. To proceed I asked several questions;

Who were the electorates?

What issues and needs were most important to them?

What conversation were these electorates having?

Where were they communicating from?

At what times or periods were they communicating?

In seeking to understand who the electorates were, one would gain insight the demography of the audience. This piece of information many times show that the young, middle aged and the old all have different things they are particular about. It clearly defines the issues that each age group or gender finds more important.

Studying the social conversations, interactions, behavioral pattern and sentiments of these different audience groups points to their concerns. This piece of information also leads you to individuals or organizations influencing the conversations, a key insight when optimizing your content.

The devices used by your audience is a key metric. The way people interact with news on mobile devices is quite different than news behavior on the desktop/laptop computers. Data has shown that people spend far more time with news apps on the smartphone and tablet, visit more pages at a sitting, and return more frequently than they do on conventional computers.

The time and periods where your audience is active, is key because you will not want to serve them content when they are asleep or serve them a particular content when they are interested in another.

How publishers are augmenting the news dissemination

Top media organizations are waging a war on your attention to see who comes out tops. Buzzfeed for example has experience exponential growth in the past few years, not by advertising but by leveraging technology - A Headline Optimization Tool for split-testing headlines. This will be discussed in detail at the Tools section of this module.

Headline is king. The Headline Optimization Tool I mentioned earlier uses a technique called A/B Headline Testing. This technique is used to determine the best headline for an article using some metric like the number of clicks on the story.

Reporters while writing a story furnish it with a number of other headlines. The news site displays those alternate versions of the headlines to different audience in order to study they click behavior.

At a point, the monitoring algorithm will pull results together to show which of the alternate headlines was clicked the most. This method can also be used for video publications and images.

Methodologies to enhance the spread of news

News outlets will not back down when it comes to finding ways to retain your attention. They will continue to invest in resources to produce methods to generate clicks. I will talk about some of them.

Widgets: At the bottom of many articles, you will a number of other articles calling out to you. The algorithm used here has studied you clicks, your reading preference and history and serves you the content it feels you most likely will want.

Question headlines: Evidence shows that headlines styled as questions lead to a different response compared to the traditional and forward-reference headlines, especially when paired with a topic the participants found particularly unappealing. Let’s take a look at these two headlines; “Breaking News: A spaceship has been spotted across Egypt” and “Breaking News: An UFO has been spotted flying past the pyramids in Egypt, do you think Aliens are coming?”. One would find that the reaction between both of them will be quite different.

Thumbnails: Flashy thumbnails and headlines are mostly there to keep readers clicking. Publishers experiment with different thumbnail images. Many times the algorithm is random.

Timeliness: Earlier I mentioned the time factor, which is key to not just creating a content that is timely, but knowing what time to publish it to your social media. This is so because your audience is interested in different things at different times.

Tools to optimize the performance of content online

Newsrooms use a wide range of tools to harness data towards creating even better content. Some of them include;

Echobox: This tool uses Artificial Intelligence to drive your social media activity. Its analytics feature collects and aggregates your social media data to help you discover which post performed best, what contents you should re-share. It also shows you what kind of content to recreate and the editors or channels that performed best.

Keyhole: I used this tool in 2014 and ’15 for the EDAV project. At the time it was still free. Keyhole is a hashtag tracking software that measures keywords and hashtags metrics in real-time. Publishers can also use this to understand the sort of content they need to create.

Chartbeat: This tool allows you see how your audience connects with your content in real time, adapts user experience and test page elements like headlines within your normal workflow, to know what moves your most valuable audience.

NewsWhip: This tool tracks the spread of any content within seconds of publication, and predicts what will make an impact every day. It monitors trends, topics, formats, authors, publishers, social accounts, and posts across seven social platforms.

Nelio A/B Testing: This is the tool I discussed about earlier used to create alternate headlines.

Every news organization has different values, and what metric matters to them at one point or another. The trick is to measure everything, but focus on optimizing for one or two. It could be subscription rate, click rate, engagement, bounce rate or overall satisfaction.

Metric will continue to be used towards optimizing, creating and distributing news content to push readers to ask for more of the content that they want. News organizations will continue to mine troves of data littered with behavioral traits, personal characteristics, sentiments and in some cases key words posted by users on social media towards the creation of the user-friendly premium content.

In other words, you can't judge all content by the same yardstick or by the same metric. You have to think about what the content is supposed to do for your business or for your editorial strategy, and then measure the metric that helps you optimize for that.

In the next video, we will discuss the core strength of algorithms in news under the module - Platform Power and Algorithms. That will be all for now.