Hall | Welcome & instructions

Hello, welcome to my home!

As I know you have limited time to complete your
mission, I can help if I give you a few pointers.

HOW TO
PLAY?
THEMES

This is the plan
of the house with
the living rooms
which you will
have to pass
through...

HOW TO PLAY?
Goal

Have you ever heard of Escape Rooms?

Your goal is to escape the house with the certainty
that you'll bring with you the necessary knowledge to
understand the significance of algorithmic bias and its
potential consequences, recognise biases in AI
applications and have a sense of responsibility and
awareness towards ethical considerations in AI and
technology use.

HOW TO PLAY?
Goal

You'll have to explore every room in the house and
complete all the challenges to have the key that will
give you access to the next room. The aim is to escape
the house as quickly as possible and with the most points.

Don't' worry, there will be hints that will help you
overcome the challenges.

HOW TO PLAY?
Points and keys

There may be optional challenges and others that are
compulsory. For every challenge successfully solved,
the team earns points.

When you've solved all the compulsory challenges in a room,
you get a key that unlocks the passage to the next room.

HOW TO PLAY?
Challenges

Rooms 1 to 4 have the same structure:

1 opening message

2 compulsory challenges (each with 2 attempts)
2 optional challenges (with 1 attempt)
(if exhausted attempts without the right answer, don't earn points)

1 final message

HOW TO PLAY?
Pay attention

As you progress to the next room,
you can't go back to the previous one.

At the end of the game, you'll be able to revisit all the rooms,
but the challenges are already "closed", so you won't be able
to answer the optional challenges that you may not have done.

All the rooms in the game will have these features.

Let's see which team can explore all the rooms,
complete all the missions and escape the house as
quickly as possible with the most points!

If you have any questions about how the game works,
you have your Taskmaster to help you.

Are you ready?

click on the next button

Disclaimer: this game is designed to be played in teams.

Let's start by choosing an avatar for your team!

Are you ready? Let's go!

One of these nice people will be your team's avatar.

This will be chosen at random...

Click on the button ASSIGN to see which character is assigned to you!

Well done!

This is the avatar that represents your team and holds the key that gives you access to the next room!

Please take a screenshot or a photo to introduce this person to your Taskmaster by the agreed
communication channel.

Click on the key
to move on to the
next room!

Have you sent the screenshot to your Taskmaster?

Please send a screenshot of this screen to the Taskmaster. You can send it via the chat room or by email.

Once you've sent it, click on the button:

Click on the available room(s)

Hall

Hall
Welcome & instructions

Room 1

Room 1
Introduction To Ai And Potential Inequalities

Room 2

Room 2
How Ai Works And Algorithmic Bias

Room 3

Room 3
Algorithmic Bias And Its Dangers

Room 4

Room 4
Ethical Considerations In Ai: Policy Agendas And Impact In The World

Exit
Room

Exit room
Final messages

Room 1 | Introduction To Ai And Potential Inequalities

Mandatory

Mandatory

Optional

Optional

Welcome to the pink room!

Here we will explore topics such as:

- What is Artificial Intelligence (AI)?
- How does AI impact our lives?
- What Inequalities can AI Introduce?

Did you know...

AI isn't just for scientists! Did you know that the movies you watch, the games you play, and even the music you listen to are often designed or chosen by AI? For example, AI helps Netflix suggest movies you might like and Spotify finds new songs based on your listening habits. Also, AI is behind the cool features of video games such as non-player characters (NPCs) behaving like real humans.

Before exploring this room, you should take a look at this document:

In this room you have some challenges that will help you understand these themes better. All you must do is find the objects that give you access to the challenges. Let's move on and look for more information in the objects in the room?

Pppsssst, a hint...
The room can be rotated!

Charlie

Hi, how are you?

Tell me something, to what extent do you think Artificial Intelligence is already integrated into your daily life?

Hey, don't answer until you've seen this document I'm sending you...

Click to open file.

Which answer would you give?

Select the correct option.

Me: For example, by personalizing my social media content based on my browsing habits and preferences.

Me: AI has no current applications in my everyday life; it might become relevant in the future.

Well done! Your answer is correct. Check the additional information.

Your answer is not correct. Check the additional information.

Additional information

AI is not a future technology waiting to be implemented; it’s already deeply embedded in various aspects of daily life including social media.

For instance, AI is used in navigation systems like GPS, voice assistants such as Siri and Alexa, recommendation systems on streaming platforms and even in optimizing traffic flows in cities. It is used across numerous sectors and everyday technologies.

Additional information

AI is not a future technolgoy waiting to be implemented; it’s already deeply embedded in various aspects of your daily life. In case you are not using social media, AI is present in many other ways is your life.

For instance, AI is used in navigation systems like GPS, voice assistants such as Siri and Alexa, recommendation systems on streaming platform and even in optimizing traffic flows in cities. Claiming that AI might only be used in the future ignores its current widespread applications across numerous sectors and everyday technologies.

Charlie

Hello, is everything all right?

Tell me, have you ever wondered why does your friend see different content on her social media app compared to what you see?

Select the correct option.

Me: Their content is personalized based on their individual preferences, interactions, and browsing history, which are likely different from yours.

Me: Social media content appears randomly for each user without any specific personalization or pattern.

Well done! Your answer is correct. Check the additional information.

Your answer is not correct. Check the additional information.

Additional information

Social media platforms use sophisticated AI algorithms to analyze individual user behavior, including likes, shares, comments, and time spent on specific posts.

This data is used to create a personalized feed that caters to each user's preferences, ensuring that the content displayed is relevant to their interests.

This process increases user engagement by showing content that the individual is more likely to enjoy and interact with.

Additional information

Social media platforms invest heavily in algorithmic personalization to maximize user engagement and satisfaction.

Random distribution would not serve this purpose as it would likely lead to irrelevant content delivery, which could decrease user engagement and satisfaction with the platform.

This misrepresentation overlooks the complex and purposeful design behind content curation on social media.

Are AI systems inherently free from inequalities?

Select the correct option:

A

Yes, AI systems are designed with advanced algorithms that automatically correct any biases in the data.

B

Yes, since AI systems process data mathematically.

C

No, if they are trained with biased data or implemented without careful consideration of their social impacts.

Well done! Your answer is correct.
AI systems, like any technology that relies on data, are susceptible to the biases present in that data. Since these systems learn from historical data, any existing prejudices or disparities in the data can lead to biased outcomes. Additionally, the design and deployment of AI can inadvertently introduce new biases if not carefully managed.

Your answer is incorrect.
AI algorithms can identify and mitigate some types of bias but cannot automatically correct all biases in the training data. Effective bias mitigation requires ongoing human oversight, system updates, and adaptations based on new understandings of biases. Algorithms and models are built by humans and trained with human-generated data, which can carry implicit biases. Therefore, without deliberate interventions, AI systems are not inherently free from inequalities.

Can you trust that all AI systems around you are always beneficial for you and the environment?

Select the correct option:

A

No, not all AI systems are guaranteed to be beneficial as their impact depends on how they are designed, the data they use, and the purposes for which they are deployed.

B

No, because all AI systems consume a significant amount of energy, they inherently harm the environment, regardless of their intended use or benefits.

C

Yes, all AI systems are programmed to automatically benefit both individuals and the environment without any negative consequences.

Well done! Your answer is correct.
AI systems' benefits and drawbacks depend on nuanced factors including their design, implementation, and the integrity of their data. Ethical considerations and potential for harm need to be evaluated based on specific circumstances and the safeguards put in place.

Your answer is incorrect.
AI systems can have unintended consequences if not properly designed or if their deployment does not consider broader societal and environmental impacts. For instance, AI-driven processes might increase efficiency but also lead to environmental challenges if not managed responsibly. Although some AI systems are energy-intensive, many are designed for energy efficiency and aim to reduce environmental impact, highlighting the diversity and potential sustainability benefits of AI applications.

Which painting is a created with AI (DALL-E)?

Select the correct option:

Sunflower fields

Is this a copy of the painting by a famous European artist?

Girl with a pearl earring

Is this a copy of the painting by a famous European artist?

Your answer is correct.
”Sunflower fields” was input into the AI tool DALL-E to generate a painting inspired by Vincent Van Gogh. While AI is capable of producing art, this practice leads to concerns about copyright and the ethical implications of digital replication, especially when AI mimics established artists.
What are your thoughts on the ethics of co-creating with machines that replicate the works of real artists?

Your answer is incorrect.
"Girl with a pearl earring" is a renowned artwork by Johannes Vermeer, a master of the Dutch Golden Age. Created around 1665, this painting is celebrated for its subtle use of light and color and the intimate and introspective depiction of the subject, making it one of the most beloved pieces in art history.
Picture from Wikipedia.

Want to see a clue? Click here

Please match the following statements around Artificial intelligence:

Make the correct match and click SUBMIT.

AI refers to the development of...
AI is...
If not carefully designed AI may cause...
... computer systems that can perform tasks typically requiring human intelligence.
... embedded in our daily lives.
... unequal treatment across demographics.

Well done! Your answer is correct.
1. AI refers to the development of computer systems that can perform tasks typically requiring human intelligence. This correctly defines what AI is.
2. AI is embedded in our daily lives, accurately reflecting how integrated AI has become in various aspects of our everyday routines.
3. If not carefully designed, AI may cause unequal treatment across demographics, highlighting the importance of ethical considerations in AI development.

Your answer is incorrect.
Let's review the matches! It looks like there was a mix-up:
1. AI refers to the development of computer systems that can perform tasks typically requiring human intelligence. This statement explains what AI fundamentally is.
2. AI is embedded in our daily lives, indicating its pervasive presence and usage.
3. If not carefully designed, AI may cause unequal treatment across demographics, highlighting the potential ethical and societal risks associated with AI.

Solution

Match of the following statements around Artificial intelligence.

AI refers to the development of...
AI is...
If not carefully designed AI may cause...
... computer systems that can perform tasks typically requiring human intelligence.
... embedded in our daily lives.
... unequal treatment across demographics.

Dolor eu fermentum dui. Proin quis lorem lacus. Aliquam erat volutpat. Phasellus vulputate dolor gravida porttitor molestie. Sed pulvinar diam ut eleifend commodo. Maecenas a faucibus tortor, id semper arcu. Duis eget varius massa, ut faucibus?

Draw, take a photo and send it to your taskmaster.

 

Once you’ve sent your photo, click on the "Challenge completed" button.

Well done!
Mission completed!

 

Thank’s for sharing your work!

Want to see a clue? Click here

Room completed

You've already completed all the compulsory challenges.

Have you already done the 2 optional challenges? If you haven't, you can choose to stay in this room and do them. Or you can move on to the next room.

The choice is yours!

You've just taken the first steps into the fascinating world of Artificial Intelligence (AI)!

By now, you understand what AI is, how it's changing our lives, and some challenges it brings, including potential inequalities.

Always remember, while AI can make our lives easier and more fun, it's essential to use it wisely and think about how it affects everyone around us.

Stay curious, question the norms, and use what you've learned today to help build a fairer future with technology.

Ready to see what's next? Let's dive deeper!

Now, take the key and go to the next room!

Click on the available room(s)

Hall

Hall
Welcome & instructions

Room 1

Room 1
Introduction To Ai And Potential Inequalities

Room 2

Room 2
How Ai Works And Algorithmic Bias

Room 3

Room 3
Algorithmic Bias And Its Dangers

Room 4

Room 4
Ethical Considerations In Ai: Policy Agendas And Impact In The World

Exit
Room

Exit room
Final messages

Room 2 | How Ai Works And Algorithmic Bias

Mandatory

Mandatory

Optional

Optional

Welcome to the yellow room!

Here we will explore topics such as:

- How AI learns: from data, using neural networks, through training and testing
- Different types of bias in AI: data bias, algorithmic bias and societal bias.
- The importance of fairness in AI: ensuring fair decisions, understanding impacts and ethical development.

Did you know...

AI systems can be trained to recognize not only fruits but also faces, animals, and even handwritten notes!
However, if the AI isn't trained with diverse data, it might think a chihuahua is a blueberry muffin! (Yes, it really happened!) Plus, some AI can even create art and music that looks and sounds like it was made by humans, and others are used in video games to create smarter opponents!

Before exploring this room, you should take a look at this document:

In this room you have some challenges that will help you understand these themes better. All you must do is find the objects that give you access to the challenges. Let's move on and look for more information in the objects in the room?

Match the types of biases with their descriptions.

Make the correct match and click SUBMIT.

Data Bias
Algorithmic or Model Bias
Human-driven or Societal Bias
This happens when the training data isn’t diverse enough. Example: If an AI is trained only on pictures of one kind of dog, like beagles, it may not recognize other breeds.
Occurs when the AI model accidentally prefers certain groups. Example: If a hiring algorithm favors resumes that have certain keywords, it might overlook qualified candidates who use different words.
This means that AI reflects unfair attitudes from society. Example: If an AI used in law enforcement is trained on biased past crime data, it may unfairly target certain communities.

Well done! Your answer is correct.
You matched each type of bias with its correct description and example.
Understanding these distinctions helps identify and address biases in AI systems effectively.

Your answer is incorrect.
Remember, data bias stems from limited training data, algorithmic bias comes from model preferences, and societal bias reflects societal attitudes.

Solution

Match the types of biases with their descriptions.

Data Bias
Algorithmic or Model Bias
Human-driven or Societal Bias
This happens when the training data isn’t diverse enough. Example: If an AI is trained only on pictures of one kind of dog, like beagles, it may not recognize other breeds.
Occurs when the AI model accidentally prefers certain groups. Example: If a hiring algorithm favors resumes that have certain keywords, it might overlook qualified candidates who use different words.
This means that AI reflects unfair attitudes from society. Example: If an AI used in law enforcement is trained on biased past crime data, it may unfairly target certain communities.
Want to see a clue? Click here

Drag each AI application to the correct description.

Make the correct match and click SUBMIT.

Healthcare Diagnosis
Autonomous Vehicles
Recommendation Systems
Uses AI to analyze medical images and patient data to diagnose diseases and recommend treatments.
Uses sensors and AI to navigate roads, avoid obstacles, and make driving decisions.
Uses user data and AI to suggest products, movies, or music that a user might like.

Well done! Your answer is correct.
Excellent work! You correctly matched each AI application with its description. Your understanding of AI applications in healthcare, autonomous vehicles, and recommendation systems is spot on!

Your answer is incorrect.
Almost there! Check your matches again. Remember, Healthcare Diagnosis involves analyzing medical images and patient data, Autonomous Vehicles navigate roads using AI, and Recommendation Systems suggest products based on user data.

Solution

Match the types of biases with their descriptions.

Healthcare Diagnosis
Autonomous Vehicles
Recommendation Systems
Uses AI to analyze medical images and patient data to diagnose diseases and recommend treatments.
Uses sensors and AI to navigate roads, avoid obstacles, and make driving decisions.
Uses user data and AI to suggest products, movies, or music that a user might like.
Want to see a clue? Click here

Charlie

Hi, I was just thinking that the way AI learns from data is similar to the way we teach a pet parrot to recognise fruit :D What do you think? I'll give you two hypotheses:

First one: the parrot (AI) is programmed with fixed information about fruits, uses pre-defined rules, and doesn’t need to see new fruits.

Or… the parrot (AI) is shown many pictures of different fruits (data), learns patterns (features) and identifies new fruits (testing).

How would you answer?

Select the correct option.

Me: Ah, ah, ah... what a nice thought! I think it's your second hypothesis.

Me: Hum, funny you should think of that, interesting! I think it's the first hypothesis you mentioned.

Well done! Your answer is correct. Check the additional information.

Your answer is not correct. Check the additional information.

Additional information

You correctly explained how AI learns by drawing an analogy to a pet parrot.

The parrot (AI) learns by seeing many examples of fruits (data). It identifies patterns like shape and color (features), then recognizes new fruits (testing).

Additional information

Remember, AI (like the parrot) learns from examples (data), identifies important features (like shape and color), and tests its learning by recognizing new fruits.

For instance, if enough pictures are shown, the parrot can tell apples from oranges based on these features.

Charlie

Hi! How are you? Have you ever thought that neural networks process information in a similar way to brain cells?

Tell me your opinion... do you think that 1- neural networks are like independent brain cells - they follow fixed rules and don’t need data to make decisions... Or…

2- Neural networks are like interconnected brain cells - they process information and learn from data?

How would you answer?

Select the correct option.

Me:Interesting, yes! ... I agree with your second hypothesis.

Me:Oh yes, I see what you mean, I think it's your first hypothesis!

Well done! Your answer is correct. Check the additional information.

Your answer is not correct. Check the additional information.

Additional information

You described neural networks well.

They are like interconnected brain cells. They process information, make data-based decisions, and learn from experiences.

Additional information

Consider neural networks as interconnected brain cells that process information and make decisions based on data.

Just like brain cells learn from experiences, neural networks improve by learning from data. For example, they might learn to recognize images of animals over time by processing many pictures.

What is data bias in AI?

Select the correct option:

A

It shows when an AI model unintentionally favors certain groups.

B

Data bias occurs when training data isn't diverse enough.

C

Both answers are correct.

Well done! Your answer is correct.
Both statements are accurate.
Data bias in AI occurs when a model unintentionally favors certain groups, often due to non-diverse training data.

Your answer is incorrect.
The right answer is C - “Both answers are correct”.
Data bias arises when a model unintentionally favors certain groups, often due to insufficient diversity in the training data.

Want to see a clue? Click here

Which photo best represents a diverse training dataset for AI learning?

Select the correct option:

Photo A

This photo shows a basket containing a mix of fruits, including apples, bananas, and oranges.

IMAGE SOURCE: image created using Dall-E

Photo B

This photo shows a basket filled only with green apples.

IMAGE SOURCE: image created using Dall-E

Your answer is correct.
Great choice! Photo A, showing a variety of fruits, represents a diverse training dataset. Just like learning about different fruits helps you recognize them all, AI needs varied data to understand and perform well in different situations.

Your answer is incorrect.
Not quite. Photo B, showing only green apples, lacks diversity. For AI to learn effectively, it needs a variety of examples. Imagine trying to recognize all fruits by only seeing green apples; you’d miss out on identifying other fruits. AI works the same way.

Want to see a clue? Click here

Dolor eu fermentum dui. Proin quis lorem lacus. Aliquam erat volutpat. Phasellus vulputate dolor gravida porttitor molestie. Sed pulvinar diam ut eleifend commodo. Maecenas a faucibus tortor, id semper arcu. Duis eget varius massa, ut faucibus?

Draw, take a photo and send it to your taskmaster.

 

Once you’ve sent your photo, click on the "Challenge completed" button.

Well done!
Mission completed!

 

Thank’s for sharing your work!

Want to see a clue? Click here

Room completed

You've already completed all the compulsory challenges.

Have you already done the 2 optional challenges? If you haven't, you can choose to stay in this room and do them. Or you can move on to the next room.

The choice is yours!

In this room you've successfully covered the basic mechanics of artificial intelligence (AI) systems and the concept of algorithmic biases.

AI learns through data analysis, similar to teaching a parrot to recognize fruits from images. Neural networks serve as AI's foundation, processing information and making decisions based on patterns in data.

The training and testing phases are crucial in AI development, where the system learns from datasets to identify key features and predict outcomes for new data. However, the text emphasizes the risk of algorithmic biases creeping into AI systems. These biases can stem from limited diversity in training data, inherent biases in the AI model itself, or societal biases that influence its development.

The consequences of biased AI are significant, especially in applications such as job recruitment, where unfair decisions can perpetuate inequalities. The passage stresses the responsibility of developers to mitigate biases and ensure AI systems operate fairly in various aspects of life.

Now, take the key and go to the next room!

Click on the available room(s)

Hall

Hall
Welcome & instructions

Room 1

Room 1
Introduction To Ai And Potential Inequalities

Room 2

Room 2
How Ai Works And Algorithmic Bias

Room 3

Room 3
Algorithmic Bias And Its Dangers

Room 4

Room 4
Ethical Considerations In Ai: Policy Agendas And Impact In The World

Exit
Room

Exit room
Final messages

Room 3 | Algorithmic Bias And Its Dangers

Mandatory

Mandatory

Optional

Optional

Welcome to the blue room!

Did you know...
that algorithms and artificial intelligence can sometimes make unfair decisions?

This happens due to different types of bias, like data-driven bias, model-driven bias, and human-driven bias. But don’t worry, there are ways to fix this!

In this room we will explore topics such as:

- Data-driven bias: What is it? Examples and scenarios.
- Algorithmic or model-driven Bias: What is it? Examples and scenarios.
- Human-driven or societal bias: What is it? Examples and scenarios.

We will also learn the different strategies that can be used to mitigate this situation.

Before exploring this room, you should take a look at this document:

In this room you have some challenges that will help you understand these topics better. All you must do is find the objects that give you access to the challenges. Let's move on and look for more information in the objects in the room.

If you needed to upload an image of professional working in a hospital setting, which picture would you select?

Select the correct option:

Professional team working in a hospital

IMAGE SOURCE: image created using Dall-E

Professional team working in a hospital

IMAGE SOURCE: image created using Dall-E

Yes, the best choice!
Make sure you include images with a balanced mix of genders in different positions to avoid bias.

Pay attention!
When creating or training AI, make sure the data you use includes diverse representations of people in various professional roles. In this picture, the image shows only male doctors and female nurses: it reinforces biases. Instead, include images with a balanced mix of genders in different positions to avoid bias.

Want to see a clue? Click here

If you needed to upload an image of IT professionals, which picture would you select?

Select the correct option:

IT professionals

IMAGE SOURCE: image created using Dall-E

IT professionals

IMAGE SOURCE: image created using Dall-E

Well done! Your answer is correct.
By including a diverse range of individuals, AI models gain a more comprehensive understanding of human experiences, mitigating the risk of biased outcomes and fostering greater fairness and equity in decision-making processes.

Your answer is incorrect.
When crafting or training AI systems, prioritize datasets that encompass a diverse spectrum of individuals across different professional domains. For instance, in the provided image, featuring solely male IT professionals perpetuates biases and limits the inclusivity of the dataset. Instead, opt for images showcasing a balanced representation of genders within various professional roles.

Want to see a clue? Click here

Match the correct statement into the box that best represents a danger of bias in AI systems.

Make the correct match and click SUBMIT.

Perpetuating Inequalities
Data Issues
Unfair Treatment
AI systems can perpetuate existing social inequalities by reinforcing stereotypes and discriminatory practices that are already present in society.
AI systems can make decisions based on incomplete or unrepresentative data, leading to inaccurate outcomes that do not reflect the diversity of the population.
AI systems can lead to unfair treatment of certain groups by disproportionately affecting marginalized communities and exacerbating existing disparities.

Well done! Your answer is correct.
Fantastic job! You’ve accurately matched the statements to their respective boxes: perpetuating inequalities, data issues, and unfair treatment.
By understanding how AI systems can perpetuate social inequalities, make decisions based on incomplete or unrepresentative data, and lead to unfair treatment of certain groups, you’re showcasing a comprehensive grasp of the dangers of bias in AI.
This knowledge is crucial for developing and using AI responsibly. Keep up the excellent work and continue to delve into the ethical implications of technology to ensure a fair and equitable future for all. Well done!

Your answer is incorrect.
Don’t worry, it’s a tricky topic! It looks like some of the statements didn’t match the correct boxes.
Remember, AI systems can perpetuate social inequalities, make decisions based on incomplete data, and lead to unfair treatment of certain groups.
Take another look at the statements and try again. Understanding these dangers is key to using AI responsibly.
Keep at it, you’re doing great!

Solution

Match the correct statement into the box that best represents a danger of bias in AI systems.

Perpetuating Inequalities
Data Issues
Unfair Treatment
AI systems can perpetuate existing social inequalities by reinforcing stereotypes and discriminatory practices that are already present in society.
AI systems can make decisions based on incomplete or unrepresentative data, leading to inaccurate outcomes that do not reflect the diversity of the population.
AI systems can lead to unfair treatment of certain groups by disproportionately affecting marginalized communities and exacerbating existing disparities.
Want to see a clue? Click here

Match the type of Algorithmic bias with its example:

Make the correct match and click SUBMIT.

Unfair Treatment in Hiring
Biased Predictive Policing
Discriminatory Loan Approvals
Example: AI systems that prioritize candidates from certain demographics due to biased training data
Example: AI algorithms that unfairly target specific communities based on historical data.
Example: AI systems denying loans to individuals based on factors unrelated to their creditworthiness.

Well done! Your answer is correct.
Congratulations on correctly identifying each danger of AI and its corresponding example!:
Unfair Hiring Practices: This happens when AI systems favor candidates from certain groups because of biased training data.
Biased Predictive Policing: This occurs when AI unfairly targets specific communities based on biased historical data.
Discriminatory Loan Denials: This refers to AI systems rejecting loans for reasons unrelated to creditworthiness, which perpetuates unfair lending practices.

Your answer is incorrect.
Ensure accurate matching by associating each danger of AI with its appropriate example:
Unfair Treatment in Hiring: this refers to AI systems favoring candidates from certain demographics due to biased training data, potentially leading to unfair hiring practices.
Biased Predictive Policing: this involves AI algorithms unfairly targeting specific communities based on biased historical data, influencing discriminatory policing practices.
Discriminatory Loan Approvals: This relates to AI systems denying loans to individuals based on factors unrelated to their creditworthiness, perpetuating discriminatory lending practices.

Solution

Match the type of Algorithmic bias with its example:

Unfair Treatment in Hiring
Biased Predictive Policing
Discriminatory Loan Approvals
Example: AI systems that prioritize candidates from certain demographics due to biased training data
Example: AI algorithms that unfairly target specific communities based on historical data.
Example: AI systems denying loans to individuals based on factors unrelated to their creditworthiness.
Want to see a clue? Click here

Charlie

Hey everyone, I've been reading up on AI ethics, and I'm a bit concerned about biases in AI systems, especially with ChatGPT being so widely used.

Please read this information I'm sending you... And tell me what do you think of this.

Click to open file.

Which answer would you give?

Select the correct option.

Me: I get where you're coming from, but honestly, I think we're blowing this AI bias thing out of proportion.

Me: Absolutely, but being aware of this is key. We can take steps to ensure that ChatGPT is used responsibly and that biases are minimized in its applications.

Well done! Your answer is correct. Check the additional information.

Your answer is not correct. Check the additional information.

Additional information

Absolutely, acknowledging the widespread use of ChatGPT and being concerned about biases are valid points.
It's crucial to recognize that biases in AI systems can have real-world implications. By understanding and actively addressing these biases, we can ensure that ChatGPT and similar technologies are used responsibly and fairly. Strategies like using diverse and fair data, questioning biased results, understanding AI decision-making, and advocating for fairness are essential steps toward mitigating biases in AI applications.
Your awareness and proactive stance contribute to the ethical use of AI.

Additional information

Your perspective suggests that you might not fully grasp the significance of AI biases.
The concern about biases in AI systems, including ChatGPT, is not exaggerated but rather a critical issue. Biases can lead to unfair outcomes, perpetuate inequalities, and affect various aspects of life where AI is applied, such as job recruitment or healthcare decisions.
Understanding and addressing biases are essential for ensuring that AI systems like ChatGPT are used ethically and responsibly.

What is Algorithmic Bias?

Select the correct option:

A

Algorithmic bias refers to the systemic and repeatable errors in computer systems that always lead to accurate outcomes.

B

Algorithmic bias refers to the systemic and repeatable errors in computer systems that lead to unfair outcomes.

C

Algorithmic bias refers to the random errors in computer systems that sometimes lead to unfair outcomes.

Well done! Your answer is correct.
The correct answer highlights that algorithmic bias pertains to consistent and systematic errors within computer systems, resulting in unfair outcomes. This bias can occur due to various factors, such as biased data used for training, flawed algorithms, or discriminatory decision-making processes. Importantly, these errors are not isolated incidents but rather recur systematically, leading to persistent inequities in AI systems.

Your answer is incorrect.
Algorithmic bias consistently leads to unfair outcomes, not accurate ones, which undermines the fairness and integrity of computer systems. These biases can arise from factors such as biased data, flawed algorithms, or improper implementation. It's crucial to recognize that algorithmic bias is not sporadic or random but rather characterized by systemic and repeatable errors in computer systems. These errors persistently result in unjust treatment or discrimination against certain groups, deeply impacting the design, training, or implementation of algorithms.

Want to see a clue? Click here

Dolor eu fermentum dui. Proin quis lorem lacus. Aliquam erat volutpat. Phasellus vulputate dolor gravida porttitor molestie. Sed pulvinar diam ut eleifend commodo. Maecenas a faucibus tortor, id semper arcu. Duis eget varius massa, ut faucibus?

Draw, take a photo and send it to your taskmaster.

 

Once you’ve sent your photo, click on the "Challenge completed" button.

Well done!
Mission completed!

 

Thank’s for sharing your work!

Want to see a clue? Click here

Room completed

You've already completed all the compulsory challenges.

Have you already done the 2 optional challenges? If you haven't, you can choose to stay in this room and do them. Or you can move on to the next room.

The choice is yours!

Algorithmic bias happens when computers make unfair choices again and again. For this reason, it is very important to understand this as AI and ML become more common.

Remember, there are different types of algorithmic bias:
- Data-Driven Bias occurs when computers learn unfair ideas from the data they use, like an AI hiring system favouring men for certain jobs.
- Algorithmic or Model-Driven Bias happens accidentally during AI model building, such as a facial recognition system being better at recognizing light-skinned faces.
- Human-Driven or Societal Bias comes from people's biases, like unfairly labelling pictures.

Detecting bias is like solving a puzzle, and it's crucial to ensure fairness at every step of AI development, just like baking a cake with fair ingredients from start to finish. Detecting and fixing bias is tough but essential: we need diverse data, to ask questions, understand how AI works, and speak up for fairness.

Now, take the key and go to the next room!

Click on the available room(s)

Hall

Hall
Welcome & instructions

Room 1

Room 1
Introduction To Ai And Potential Inequalities

Room 2

Room 2
How Ai Works And Algorithmic Bias

Room 3

Room 3
Algorithmic Bias And Its Dangers

Room 4

Room 4
Ethical Considerations In Ai: Policy Agendas And Impact In The World

Exit
Room

Exit room
Final messages

Room 4 | Ethical Considerations In Ai: Policy Agendas And Impact In The World

Mandatory

Mandatory

Optional

Optional

Welcome to the orange room - Ethical Considerations in AI: policy agendas and impact in the world

By now, you've encountered some of the amazing things AI can do. But with great power comes great responsibility!

In this room, you'll transform into an AI Ethics Detective, tackling challenges that ensure AI is fair and benefits everyone. We'll be exploring the concepts of transparency, bias, and data privacy to make sure AI decisions are ethical and responsible.

Ready to put your detective skills to the test? Let's dive in!

Before exploring this room, you should take a look at this document

In this room you have some challenges that will help you understand these themes better. All you must do is find the objects that give you access to the challenges. Let's move on and look for more information in the objects in the room?

AI helps with loan approvals, but can be biased. What might cause this?

Select the correct option:

A

AI database gets regular data updates.

B

AI prioritizes speed over accuracy.

C

AI trained on mostly wealthy applicants.

Well done! Your answer is correct.
Nailed it!
AI trained on limited data can inherit biases. Great job identifying a potential pitfall.

Your answer is incorrect.
Think again!
Data bias can creep in through different way. Review the Room document to refresh your memory and come back when ready!

Want to see a clue? Click here

An AI raises your insurance rate, but won't explain why. Why is Transparency important?

Select the correct option:

A

To prevent the AI from getting too powerful.

B

So you can understand how the decision was made and challenge it.

C

So the AI can learn from its mistakes.

Well done! Your answer is correct.
Spot on!
Transparency allows you to understand the AI's decision-making process. This can help you identify errors or biases and potentially contest the outcome.

Your answer is incorrect.
That's not quite it.
While AI can learn and improve, transparency is more about understanding how the AI reached a decision in your specific case.

Want to see a clue? Click here

Imagine you're a legal assistant working on a new AI system for court cases. How should it be used?

Select the correct option:

Option 1 - Automatic

The AI system analyzes the case details using complex algorithms and automatically gives a decision

IMAGE SOURCE: image generated by Canva

Option 2 - Manual

The AI system analyzes the case details and creates a concise summary to be examined by the judge and the jury

IMAGE SOURCE: Helixconnect Europe

Your answer is correct.
Excellent! Option 2, where the AI provides a clear and concise summary, is the more ethical approach. Transparency is crucial in AI-assisted decision-making. This method allows the judge to understand the AI's insights alongside relevant legal precedents, enabling informed and transparent judgments.

Your answer is incorrect.
Not quite! While AI analysis can be valuable in court cases, relying solely on a complex algorithm might raise concerns about transparency. In Option 1, the judge wouldn't fully grasp the AI's calculations. Option 2, with a clear summary, fosters transparency and allows the judge to make informed decisions.

Want to see a clue? Click here

You're working as a data detective for a new fitness tracker app. The app developers are considering two ways to collect user data. Which would be the best approach?

Select the correct option:

Option 1 - Automatic

A complex app interface with various toggles and permissions for data access (e.g., location tracking, microphone access, camera access), which extracts data automatically.

IMAGE SOURCE: image generated by Canva

Option 2 - Manual

A simple and user-friendly app interface. It shows clear categories for users to input their data manually (e.g., age, height, weight, activity goals).

IMAGE SOURCE: image generated by Canva

Your answer is correct.
Great job, data detective! Option 2, where users manually input their data, is the more ethical approach. This method gives users clear control over what information they share with the app, fostering transparency and respecting user privacy.

Your answer is incorrect.
Hmm, that's not quite right. While Option 1 might collect more data for personalized fitness plans, it raises privacy concerns. Users might not understand exactly how their data (like location or microphone access) is being used, making it less transparent. Option 1 gives users more control over their data.

Want to see a clue? Click here

Match the real-life initiatives with the category that best describes their role in building a fair AI future:

Make the correct match and click SUBMIT.

Awareness Campaigns
UNESCO AI Framework
Transparency Laws
Global Standards for AI
Public Engagement
Local Regulations

Well done! Your answer is correct.
You've successfully matched the initiatives with their corresponding categories for building a fair and responsible AI future. You’re on your way to becoming an AI expert!

Your answer is incorrect.
There seems to be a mismatch in some of the pairings. Try re-reading the descriptions of the initiatives and categories in the room’ s initial message document.

Solution

Match the real-life initiatives with the category that best describes their role in building a fair AI future:

Awareness Campaigns
UNESCO AI Framework
Transparency Laws
Public Engagement
Global Standards for AI
Local Regulations
Want to see a clue? Click here

Charlie

Hey, did you hear? They’re going to start using AI to track our scores at school. I wanted to ask the teacher about it, but I’m scared she won’t like it!

I send you some information that may be useful.

Click to open file.

What would you say?

Select the correct option.

Me: You shouldn’t be scared to ask! This is a big change, so you have the right to be properly informed about how this new system works.

Me: Well, if the school decided to do it, then it should be for the best. It’s better not to complicate things.

Well done! Your answer is correct. Check the additional information.

Your answer is not correct. Check the additional information.

Additional information

That's the explorer spirit!
As a young person who cares about AI, it's important to be curious and ask questions, especially when new technologies are introduced.
Schools using AI to track scores sounds interesting, but it's also important to understand how it works.
What kind of data will it collect? How will it be used?
By asking your teacher, you're taking an active role in making sure this new system is implemented fairly and accurately.
Remember, AI is a powerful tool, and together we can make sure it's used for good!

Additional information

Uh oh, it sounds like you're letting fear hold you back.
While respecting teachers is important, you also have the right to be informed about how AI is used at school.
New technologies can be complex, and it's okay to ask questions!
Think about it this way: what if the AI tracking system makes a mistake and gives you a wrong score? Wouldn't you want to know and have it corrected?
If you don’t feel very comfortable asking too many questions to the people in charge, you can always get your friends to back you up!
Remember, as an AI explorer, you have the power to learn and speak up about ethical AI development!

Dolor eu fermentum dui. Proin quis lorem lacus. Aliquam erat volutpat. Phasellus vulputate dolor gravida porttitor molestie. Sed pulvinar diam ut eleifend commodo. Maecenas a faucibus tortor, id semper arcu. Duis eget varius massa, ut faucibus?

Draw, take a photo and send it to your taskmaster.

 

Once you’ve sent your photo, click on the "Challenge completed" button.

Well done!
Mission completed!

 

Thank’s for sharing your work!

Want to see a clue? Click here

Room completed

You've already completed all the compulsory challenges.

Have you already done the 2 optional challenges? If you haven't, you can choose to stay in this room and do them. Or you can move on to the next room.

The choice is yours!

Congratulations!

You've successfully explored the exciting world of AI and its ethical considerations.
Throughout this room, you've learned about the power of AI, its potential benefits, and the importance of responsible development. Remember:

- AI is a powerful tool that can be used to improve our lives in many ways, from education and healthcare to entertainment and scientific discovery.
- However, with great power comes great responsibility. It's crucial to ensure AI development is ethical, fair, and transparent.

You, as a young explorer, have a vital role to play in shaping the future of AI. By asking questions, staying informed, and advocating for responsible AI practices, you can help ensure AI is used for good.

Here's what you can do next:

- Keep exploring! Learn more about AI and its applications.
- Share what you've learned with your friends and family.
- Discuss the ethical implications of AI with others.
- Participate in workshops or events related to AI ethics.

Together, we can create a future where AI benefits everyone!

Now, take the key and go to the next room!

Click on the available room(s)

Hall

Hall
Welcome & instructions

Room 1

Room 1
Introduction To Ai And Potential Inequalities

Room 2

Room 2
How Ai Works And Algorithmic Bias

Room 3

Room 3
Algorithmic Bias And Its Dangers

Room 4

Room 4
Ethical Considerations In Ai: Policy Agendas And Impact In The World

Exit
Room

Exit room
Final messages

Exit room | Final messages

There you are in the exit room!

But before you leave, you have to find the four envelopes hidden in this room!

Key ideas
Room 1 | Introduction To Ai And Potential Inequalities

Dear future AI native,

Artificial Intelligence (AI) is transforming our world in incredible ways. However, it's important to remember that AI can also introduce inequalities and privacy challenges if not managed carefully.

By addressing these challenges and harnessing AI responsibly, we can create a fairer, AI-powered world.

Stay curious and informed about how AI impacts our lives!

Best, Charlie

Key ideas
Room2 | How Ai Works And Algorithmic Bias

Dear future AI native,

AI, much like a parrot learning to recognize fruits, learns from large data sets. It uses interconnected neural networks to process information and make decisions.

AI is trained on data sets and then tested on new data to evaluate its pattern recognition. However, bias can creep into AI due to data, algorithmic, and societal factors. This can lead to unfair decisions and perpetuate inequalities in various aspects of life.

Addressing this bias is a priority for developers who view it as a civil rights issue. They are working towards fairness in AI through careful development and testing.

We hope this gives you a clearer understanding of AI.

Remember, the future of technology is in your hands!

Best, Charlie

Key ideas
Room 3 | Algorithmic Bias And Its Dangers

Dear future AI native,

Algorithmic bias, leading to unfair outcomes in computer systems, often stems from systemic errors in AI and ML models.

Bias can originate from data, algorithms, or societal influences. Identifying these root causes is a complex task, much like solving a puzzle. Bias can infiltrate AI at any stage, from design to use, necessitating fairness at each step, just as in baking a cake.

Mitigating bias requires using diverse data, vigilance for biased outcomes, understanding AI decisions, and advocating for fairness. Proactive identification and addressal of bias is essential, much like a superhero fighting for digital fairness.

We hope this summary enhances your understanding of algorithmic bias.

Remember, you hold the power to shape the future of technology!

Best, Charlie

Key ideas
Room 4 | Ethical Considerations In Ai: Policy Agendas And Impact In The World

Dear future AI native,

AI is versatile, serving various purposes from predicting weather to aiding doctors and judicial systems. However, it requires ethical guidelines for responsible use.

Ethical challenges in AI include lack of transparency in decisions, risks of bias, and balancing privacy with data gathering. Tools like UNESCO's RAM ensure AI development aligns with ethical values and legal frameworks.

Global standards and recommendations for AI ethics have been established by bodies like UNESCO. The EU has enacted the EU Artificial Intelligence Act, employing a risk-based approach and banning certain practices to ensure safe AI that respects fundamental rights.

Europe emphasizes trust, transparency, and accountability in AI development to address societal challenges while safeguarding privacy and ethical standards.

We hope this summary enhances your understanding of AI ethics.

Remember, you are the future of technology!

Best, Charlie

Well done!

I'm sure it was useful to be reminded of these key ideas.

I have great news! I've just opened all the room doors so you can revisit the one you want.
Just go to the menu via this icon and click on the room you want to go to!

To move on and see the results of the game, follow me!

It’s done!

You reached the last room with

0000 points

Call your Taskmaster, only this person can help you get out of the house now and reveal the winning team of this game!

Meanwhile, you should send a screenshot of this screen to your Taskmaster.

And remember that you can revisit the rooms through the menu.

This game was developed as part of the Charlie Project, in partnership with the following organisations:

Hall

Hall
Welcome & instructions

Room 1

Room 1
Introduction To Ai And Potential Inequalities

Room 2

Room 2
How Ai Works And Algorithmic Bias

Room 3

Room 3
Algorithmic Bias And Its Dangers

Room 4

Room 4
Ethical Considerations In Ai: Policy Agendas And Impact In The World

Exit
Room

Exit room
Final messages