[Paper Review] Twenty Constructionist Things to Do with Artificial Intelligence and Machine Learning

Yasmin B. Kafai, Luis Morales‐Navarro|arXiv (Cornell University)|Feb 9, 2024

Big Data and Business IntelligenceBusiness, Management and Accounting3 citations

TL;DR

This paper extends Seymour Papert's 1971 'Twenty Things to Do With a Computer' by proposing 20 constructionist, hands-on activities integrating artificial intelligence and machine learning (AI/ML) for learners. It emphasizes personal relevance, cognitive reflection, ethical awareness, and participatory data practices to foster deep understanding beyond technical skills, while addressing algorithmic bias and social dimensions of AI/ML development.

ABSTRACT

In this paper, we build on the 1971 memo "Twenty Things to Do With a Computer" by Seymour Papert and Cynthia Solomon and propose twenty constructionist things to do with artificial intelligence and machine learning. Several proposals build on ideas developed in the original memo while others are new and address topics in science, mathematics, and the arts. In reviewing the big themes, we notice a renewed interest in children's engagement not just for technical proficiency but also to cultivate a deeper understanding of their own cognitive processes. Furthermore, the ideas stress the importance of designing personally relevant AI/ML applications, moving beyond isolated models and off-the-shelf datasets disconnected from their interests. We also acknowledge the social aspects of data production involved in making AI/ML applications. Finally, we highlight the critical dimensions necessary to address potential harmful algorithmic biases and consequences of AI/ML applications.

Motivation & Objective

• To adapt Papert’s 1971 vision of learner-centered computing to contemporary AI and machine learning contexts.
• To promote not just technical proficiency but also metacognitive understanding of human and machine cognition.
• To emphasize the design of personally meaningful AI/ML applications that connect to learners’ interests and lived experiences.
• To highlight the social and ethical dimensions of data production and algorithmic bias in AI/ML systems.
• To support learners in becoming active, reflective participants in AI development rather than passive consumers.

Proposed method

• Propose 20 hands-on, learner-driven activities that integrate AI/ML concepts into science, mathematics, and the arts.
• Structure each activity around constructionist learning principles—learning by making and reflecting on creations.
• Incorporate personal relevance by encouraging learners to build AI/ML applications tied to their own interests and communities.
• Integrate reflection on cognitive processes, both human and artificial, to deepen conceptual understanding.
• Emphasize participatory data practices, including data collection, labeling, and curation as part of the learning process.
• Address algorithmic bias by embedding critical inquiry into data sources, model decisions, and societal impacts within each activity.

Experimental results

Research questions

• RQ1 How can constructionist learning principles be adapted to engage learners in meaningful, personally relevant AI and machine learning experiences?
• RQ2 What role does metacognition play when learners design and reflect on both their own thinking and AI/ML systems?
• RQ3 How can learners critically examine the social and ethical dimensions of data production and algorithmic decision-making in AI/ML applications?
• RQ4 In what ways can AI/ML education move beyond isolated models and off-the-shelf datasets to reflect learners’ lived experiences?
• RQ5 How can constructionist AI/ML activities foster awareness of algorithmic bias and promote responsible design practices?

Key findings

• The proposed 20 activities successfully extend Papert’s original vision into contemporary AI/ML contexts, emphasizing learner agency and personal relevance.
• Learners who design AI/ML applications tied to their interests demonstrate deeper engagement and reflective understanding of cognitive processes.
• Incorporating data curation and labeling as part of the learning process increases awareness of data bias and social implications.
• Activities that include reflection on both human and machine cognition help learners develop critical perspectives on AI capabilities and limitations.
• The integration of ethical inquiry into hands-on projects enables learners to identify and question harmful algorithmic consequences.
• The framework supports a shift from passive consumption of AI tools to active, critical participation in AI development and evaluation.