Outreach

Oxford CCAI is launching its first outreach activities – an after school club to help children better talk and think about AI around them.

The club runs over 4 weeks, and in each week, researchers from the lab will use a range of activities to help children develop a better ability to talk and think about AI technologies.

  • Week 1: we will talk about AI and help children understand how AI is based on data
  • Week 2: we will help children evaluate the values that drive the design of AI
  • Week 3 & 4: we will work with children to (conceptually) design their own good AI platforms.

We publish our workshop format and material under creative commons licence. You are mostly welcome to reuse or adapt them. If you do find them useful, we would appreciate it if you could cite our work by linking our website. Also, please feel free to drop us a line if you have any feedback.

Week 1: How AI helps? + Black-box Algorithm!

Expected learning outcomes for us:

  • Understand how children perceive AI
  • Understand if children could comprehend concepts related to datafication with our instruction.

Expected learning outcomes for children:

  • Got a big idea about AI and datafication
  • Know the three aspects of datafication, i.e. data collection (privacy), datafying (algorithm), and data inference (influence)

Activity planning

  • Warm-up (Verbal discussions / Whole class: ~10 mins)

Children share their general perception of AI and some good things and not so good things about how AI can help people.

  • Introduce the drawing task (Small breakup groups: ~25 mins)

Children are provided with a bunch of scenario cards.

Each of them is provided with a piece of paper so that they can work as a group/pair, choose a scenario card, and imagine how AI may help them in these scenarios.

  • Black-box Algorithm! (Whole class: ~40-50 mins)

We used three scenarios to guide children to discuss how AI uses different data to personalise recommendations and how feedback can affect AI results.

Take home message

In the end of the session, researchers summarise the key discussion points of the day, and provide a take-home message card for children, including the following three messages:

  • AI is a smart computer that can make decisions like a human brain, and it could help us in many scenarios.
  • AI-based apps like FamilyFun can learn from the data we provide and make personalised recommendations for us.
  • AI has pros and cons. The better the data you provide, the better AI will learn, but they are still limited.

Supplementary materials

We also offer a game for children to further explore AI and algorithms. In this game, children can learn how AI and machine learning can be used to address global problems.

Week 2: Discover different values in AI!

Expected learning outcomes for us:

  • Understand children’ value on AI.
  • Help children understand three steps of AI/datafication (i.e. data collection, datafying, and influence and feedback)
  • Help children understand different values AI may contain.
  • Use a value matrix to show children how the values of different stakeholders are considered in an AI system design process.

Expected learning outcomes for children:

  • Understanding different values AI may contain.
  • Be aware of how the values of different stakeholders are considered in an AI system design process.
  • Analyse AI’s recommendation more critically.

Activity planning

  • Warm-up (Verbal discussions / Whole class: ~10 mins)

Children recall the content of the activities and their feelings from last week.

  • Different Values: Role-playing game (Group + Whole class: ~60 mins)

First, we introduce the roles and use value cards to lead the children to discuss people’s values.

Children role-play, discuss the data to be collected by the system (with data cards), and place the data cards into three piles on the data worksheet.

Then, the group exchanges the data cards with the other group, who will role-play as an AI system and assess whether sufficient data has been collected for their algorithms to function (with data worksheet).

Children continue to role-play as AI systems to make inferences and recommend ‘place of interests’ based on the data, and clarify the value/motivations of their recommendations on AI worksheet.

Children exchange the inference results and ask the original group to role-play as families to give feedback and reflect using the ‘value matrix’.

In the end, children discuss the whole class and compare their results.

Take home message

In the end of the session, researchers summarise the key discussion points of the day, and provide a take-home message card for children, including the following three messages:

  • Your data helps AI learn and suggest things based on what it has learned.
  • AI can assist with decisions, but may not understand your values. Use your own judgement too.
  • The creators’ diverse opinions can impact the suggestions AI provides

Week 3: Your own AI platform (1)

Expected learning outcomes for us:

  • Understand how children perceive ethical issues associated with AI systems.
  • Assist children in developing awareness of ethical issues and understanding the importance of reducing such issues through the consideration of AI system design.
  • Support children in creating their own AI platform with a user-centred method, i.e. begin with the user’s story.

Expected learning outcomes for children:

  • Understanding and analysing the ethical issues associated with AI systems (system bias, and chamber effects).
  • Analyse, evaluate and enhance the design of the recommendation system(FamilyFun).
  • Create an AI platform from analysing users’ needs.

Activity Planning

  • Warm-up (Verbal discussions / Whole class: ~10 mins)

Children recall the content of the activities and their feelings from last week.

  • Help AI for these challenges! (Whole class: ~10 mins)

Children learn about two cases where users complain about filter bubbles and gender bias. Children share their feelings about these phenomena and potential methods for AI platforms to address these problems.

  • AI platform/app design (Group + Whole class: ~60 mins)

We have four sections, including: 1) creating a story for the users of your AI app; 2) identifying user preferences and dislikes for this AI app; 3) translating your stories into a paper-based design of your app; and finally; 4) discussing how your app can help with Liam’s and Alice’s problems.

Children first decide the sequence in which they would like to tackle these sections for the design. Then, they go through each section. In the end, children discuss in the whole class, present their AI platform/app, and give comments on the works of other groups.

Take home message

In the end of the session, researchers summarise the key discussion points of the day, and provide a take-home message card for children, including the following three messages:

  • Design often begins with a good user story and expected actions.
  • AI systems may not always be perfect for everyone, but designers should think ahead of diverse needs and values from different users.
  • The choice of data and algorithms used by the AI systems can make a big difference and need to be fair.

Week 4: Your own AI platform (2)

Expected learning outcomes for us:

  • Assist children in developing awareness of the dual impact that an AI system has on stakeholders and understanding the importance of reducing such issues through considering AI system designs.
  • Understand how children perceive ethical issues in AI systems, especially in AI-based recommendation apps.
  • Compare and analyse the drawings in “how does AI help”.
  • Understand children’s feelings, needs and learning outcomes of this outreach programme to polish the design for further studies.

Expected learning outcomes for children:

  • Understanding and analysing AI systems’ impact on stakeholders.
  • Be aware of the potential ethical issues in recommendation systems.
  • Be aware of AI platforms limitations.
  • Analyse, evaluate and refine the design of the ethical recommendation system(FamilyFun).

Activity Planning

  • Warm-up (Verbal discussions / Whole class: ~10 mins)

Children recall the content of the activities and their feelings from last week.

  • Help AI for these challenges! (Group + Whole class: ~50 mins)

Children learn about three cases related to ethical challenges in AI-based recommendation systems: 1)‘unfair’ recommendations for different stakeholders; 2) integration with advertisements;3) consent and transparency.

Children share their feelings about these challenges and design FamilyFun to address these challenges.

  • AI platform/app design reflection (Group + Whole class: ~15 mins)

We bring back the AI platform they designed in Week 3 and their drawings of “How does AI help” in Week 1. The children modified these two works based on their new understanding of what makes a good AI platform or app.

  • Feedback (Whole class: ~15 mins)

We invite children to recall their experience in this club and share their feelings. We ask the children about their favourite and least favourite activities in this club, as well as the reasons behind their choices.

Take home message

In the end of the session, researchers summarise the key discussion points of the day, and provide a take-home message card for children, including the following three messages:

  • There are potential mistakes in AI systems and both AI designers and users should be more mindful of them.
  • AI platforms can have both good and bad impacts on users.
  • We need to understand AI’s limitations so we can use it critically.

Final reports