Ted Talk: We’re Building a Dystopia Just to Make People Click on Ads

The benefits and opportunities of machine learning (ML) algorithms translate to pioneering applications that can improve the way processes and tasks are completed. There are different ways to develop and use a machine learning system, so the benefits of a ML system are thus dependent on the way it is used for a particular purpose. However, despite its numerous benefits, there are still limitations and challenges associated with machine learning. These limitations usually revolve around the quality of data and the use given to these ML algorithms.

Download the entire lesson below in A4 PDF format:

TED Talk_ We’re building a dystopia just to make people click on ads

Warmer questions

1. Are ads intentionally targeted?
2. How exactly is advertising manipulating our spending habits?
3. To what extent is our Internet experience affected by the machine learning algorithms?
4. What would life be like in an advertising-free world?

Reading section

Battling for attention

Machine learning (ML) is a subfield of artificial intelligence that allows software applications to be more precise in predicting results. The main objective of machine learning technology is to predict an acceptable output value by building algorithms. But while these algorithms have brought important changes to the World Wide Web, advertisement and Persuasion Architecture (PA) and have also played an important role in the user’s internet experience.

Online advertising, or web advertising, is a form of marketing which uses the Internet to deliver promotional advertising messages (ads) to online consumers. Most consumers view online advertising as an unwanted interference with hardly any benefits and, for a variety of reasons, have increasingly turned to ad blocking. Many common online advertising practices are controversial and increasingly subject to regulation.

Successful PA, on the other hand, is based on the research marketers perform to map out what motivates customers through each step of their purchase experience. PA is nothing new, though. Physical retailers have used it in their stores for decades. The layout of aisles and the location of products on shelves are all prearranged to draw attention, gain interest, encourage desire, and to persuade customers to buy what they offer.

The “AIDA” Test

AIDA stands for Attention, Interest, Desire, and Action. It is a cognitive model that describes buying and selling by helping marketers appeal to the consumers’ emotional and social needs. To aid in the process of turning visitors into buyers, persuasion architecture applies AIDA to websites making the following questions:

1. WHO are we trying to persuade?
2. WHAT action do we want our customers to take?
3. HOW do we persuade our customers to take the action?

This gives rise to the questions:

1. Does the website grab our customers’ attention?
2. Does the website stimulate our customers’ interest?
3. Does the website encourage the desire to take the action of clicking deeper toward a purchase?
4. Is taking action obvious and easy?

Over the years, persuasion architecture has proven to be meaningfully effective in alluring customers and persuading them to buy their products. However, some detractors might argue, and with good reason, that PA is all about manipulation. At this point, it is worth mentioning that the difference between persuasion and manipulation lies in:

1. The intent behind the persuasion,
2. The straightforwardness and transparency of the process, and
3. The benefit or impact on individuals.

Manipulation thus implies persuasion with the intent to fool; in this case, having customers buy something that could potentially leave them either harmed or without benefit. So, are machine-learning ads and persuasion architecture targeted at manipulating us or are they just trying to persuade us into buying what is in our best interests?

Vocabulary matching

Match the vocab on the left with the correct definitions on the right.

Vocabulary gap-fill

Using the vocabulary from the previous task, fill in the missing spaces. You may need to conjugate some of the words or use a different verb or noun form.

1. Most of us are known to ____________ to persuasion through the media.
2. You may ____________ from his remarks the implications.
3. The suspects are under police ____________.
4. Many companies will only ____________ in the Sunday paper.
5. Some of the math is quite sophisticated, using differential equations, linear algebra, and covariance ____________.
6. Our special offers are intended to ____________ people to buy.
7. ____________ is used in numerous disciplines, including medical diagnosis, ad serving, spam filtering, sales forecasting, and computer vision.
8. We dug a deep hole to ____________ the animals in.
9. We will achieve much more by ____________ than by brute force.
10. Utopias are goals to be reached and ____________ goals to be avoided.
11. Make lists of what to do and ____________ your tasks.
12. It was the first commercially available machine to employ ____________.
13. The ____________ of social media users tend to fall in line more closely with those of today’s Democratic voters, for example.
14. We put an ____________ in the local paper.

TED Talk: We’re building a dystopia just to make people click on ads

We’re building an artificial intelligence-powered dystopia, one click at a time, says Techno-sociologist Zeynep Tufekci. In an eye-opening talk, she details how the same algorithms companies like Facebook, Google and Amazon use to get you to click on ads are also used to organize your access to political and social information. And the machines aren’t even the real threat. What we need to understand is how the powerful might use AI to control us — and what we can do in response.

Watch the video above and the answer the questions below:

1. What is George Orwell’s “1984”?
2. What does the speaker think we need to fear most?
3. What is being developed by companies now?
4. How many persuasion architectures can be built in the digital world?
5. What is the data that Facebook has on us?
6. Is YouTube autoplay feature a human editor?
7. What did YouTube autoplay during the rallies of then-candidate Donald Trump in 2016?
8. What did Donald Trump’s social media manager discover last year?
9. What is Facebook doing algorithmically?
10. What can these algorithms infer just from Facebook likes?
11. What can these algorithms identify?
12. What if the people in power are using these algorithms?

Advantages of using machine learning

1. Identifies trends and patterns: Machine learning algorithms can easily analyse large volumes of data and determine specific trends and patterns that would not be apparent to the human eye.
2. No human intervention needed: Since machine learning means giving machines the ability to learn, ML makes predictions and also improves the algorithms on their own. ML is also good at recognising spam.
3. Continuous improvement: As ML algorithms collect more data, they keep improving in accuracy and efficiency. This lets them make better decisions and more accurate predictions faster.
4. Handling of multi-dimensional and multi-variety data: Machine learning algorithms handle large volumes of data that are multi-dimensional and multi-variety in dynamic or uncertain environments.
5. Wide applications: ML holds the capability to help deliver a more personal experience to customers while also targeting a massive number of consumers.

Disadvantages of using machine learning

1. Data acquisition: Machine learning requires massive data that should be already unbiased and of good quality. There can also be times where algorithms must wait for new data to be created.
2. Time and resources: ML needs time for the algorithms to learn and develop enough to fulfil their purpose with a significant amount of precision and relevancy. It also needs massive resources to function. So, additional requirements of computer power and other technical supplies may be necessary.
3. Learning time: The bigger the data and the longer a machine learning system is exposed to these data, the better it will perform. ML learns through historical data to make better predictions and decisions.
4. Interpretation of results: to precisely interpret results generated by the algorithms and to carefully choose the algorithms for a certain purpose represent another challenge, too.
5. High error-susceptibility: ML is autonomous but highly susceptible to errors. Companies may end up with biased predictions coming from a biased training set. This may lead to irrelevant advertisements being displayed to customers, for example.
6. Limitations of predictions: ML systems do not understand context. Hence, depending on the given data used for training, machine learning is also vulnerable to hidden and unintentional biases. Human input is thus important to better evaluate the outputs of these machine learning systems.

Extended discussion questions

1. Can internet algorithms really predict what we want or need?
2. Or, are internet algorithms leading us towards what they “think” our likes and needs should be?
3. Are targeted ads stalking us?
4. If we can’t afford a premium internet experience, how can we stop ads?
5. Are ads really creating a parallel world for us?
6. Are we doomed to live in an everlasting advertising bubble?
7. Is persuasion architecture shaping our sense of taste?
8. Would we be lost if the ads weren’t showing us their buying options?

Debating motions

1. I don’t have time to surf the web. Ads save me a lot of time.
2. Clicking to get rid of the endless ads is dreadfully time-consuming.
3. I feel flattered and important every time I come across an ad customised with my personal info.
4. I find preposterous that advertisers use my personal data to lure me into buying their products.
5. Facebook, Google, Amazon, Alibaba, and Tencent’s machine-learning algorithms should be subject to rigorous censorship.
6. The machine-learning algorithms of companies like Facebook, Google, Amazon, Alibaba, and Tencent are helpful and beneficial for most Internet users.

Final thoughts

The collection of personal information by publishers and advertisers has also raised privacy concerns amongst Internet users. Many consumers have reservations about online behavioural predictions. Advertisers use technology to maximise their abilities to track consumers’ online activities and indiscriminately deliver ads to persuade online consumers. For some consumers, ad blocking −or ad filtering− is not sufficient enough. They already feel trapped in an architected dystopic loop filled with nonstop promotional advertising messages.

Further reading

If you’re interested in more online related topics, then I recommend visiting my lesson on how to avoid online filter bubbles using a TED Talk as a base.