- Brief created: 2024
- For policymakers
- United Kingdom (UK)
- Japan
Codesigning AI with End-Users: An AI Literacy Toolkit for Nontechnical Audiences
Based on:
Journal Article (2024) ↗
This study presents a practical, card-based toolkit designed to improve nontechnical participants’ understanding of AI concepts, empowering them to actively participate in co-design sessions for AI development.
Brief by:
Research collaborators:
Involving the public in AI design has the potential to make AI systems more transparent, ethical, and user-friendly. Yet, the limited AI knowledge among nontechnical users often leads to misunderstandings, reducing their ability to engage in meaningful design contributions. AI technologies today often operate as ”black boxes,” with end-users unable to fully understand their workings or effects. This can lead to public misconceptions about AI’s abilities, a phenomenon sometimes called the ”Superhuman Fallacy,” where AI is assumed to have capabilities it does not possess. With transparency challenges in AI-especially related to ethical issues such as bias, privacy, and accountability-these misunderstandings can become obstacles to responsible development and informed user participation.
To bridge this knowledge gap, this research presents a practical, card-based AI literacy toolkit. It aims to introduce essential AI concepts, ethical concerns, and real-world applications to nontechnical audiences in codesign settings. Each card offers a plain-language definition, relevant example, and ”What if?” prompts that encourage critical thinking, using methods grounded in human-centred design. Tested with 50 nontechnical participants, the toolkit demonstrated significant improvements in participants’ understanding of AI topics, broadening the range of AI-related ideas and critical questions raised by over 50% and increasing relevant keyword usage by 80%. Results suggest that AI literacy tools like this one can enhance nontechnical audiences’ involvement in the AI design process, fostering a more inclusive and informed approach as AI plays an increasingly impactful role in public services and digital landscapes.
Key findings
The toolkit led to significantly improved understanding and questioning of AI concepts by nontechnical users.
Evidence
Among 50 participants, those using the toolkit demonstrated a 53.94% broader range and 80.17% higher frequency of AI-related terms in their feedback compared to the control group. Terms such as ''bias,'' ''dataset,'' and ''model'' were common among toolkit users but largely absent in the control group.
What it means
This toolkit effectively equipped participants with the vocabulary and conceptual understanding needed to engage critically in AI discussions.
Toolkit use improved collaboration and mutual understanding in group settings.
Evidence
During codesign sessions, nontechnical participants reported that the toolkit ''leveled the playing field,'' establishing a shared language for discussing AI ideas. Notably, it reduced discomfort by guiding discussion when gaps arose and created a common understanding of technical terms, smoothing interactions between technical and nontechnical participants.
What it means
The toolkit facilitated a shared language and mental model among participants, essential for bridging knowledge gaps in multidisciplinary teams.
The toolkit supported more diverse and imaginative ideation around AI's applications and impacts.
Evidence
Using the toolkit's ''What if?'' prompts, participants explored various outcomes, including risks and ethical considerations. For example, one participant reflected that ''more negatives than benefits'' emerged as they discussed potential harms and benefits.
What it means
The toolkit effectively broadened participants' perspectives, encouraging them to consider both positive and negative AI impacts, crucial for responsible AI design.
Nontechnical participants valued the chance to contribute their perspectives on AI design.
Evidence
Participants expressed appreciation for the toolkit's ability to facilitate their unique insights, with two expressing a desire for ongoing input to ensure their concerns, such as ethical implications, are addressed. One participant said they valued providing a ''human perspective'' in a typically technical design process.
What it means
Nontechnical participants valued their inclusion, indicating a strong desire to advocate for public interests and contribute to ethical technology development.
Proposed action
Practitioners need to actively incorporate more diverse perspectives in AI development, including those of non-technical and underrepresented groups. This approach challenges the biases of traditional user-centred design, ensuring that diverse perspectives are represented and fostering more equitable, participatory, and just AI systems.
Comments
You must log in to ask a question
Are you a researcher looking to make a real-world impact? Join Acume and transform your research into a practical summary.
Already have an account? Log in
Discover more
Sponsored links
Codesigning AI with End-Users: An AI Literacy Toolkit for Nontechnical Audiences
Cite this brief: Mougenot, Céline. 'Codesigning AI with End-Users: An AI Literacy Toolkit for Nontechnical Audiences'. Acume. https://www.acume.org/r/codesigning-ai-with-end-users-an-ai-literacy-toolkit-for-nontechnical-audiences/
Brief created by: Dr Céline Mougenot | Year brief made: 2024
Original research:
- Smith, F., Mougenot, C., & et al., ‘Codesigning AI with End-Users: An AI Literacy Toolkit for Nontechnical Audiences’ Interacting with Computers (pp. 1–13) https://doi.org/10.1093/iwc/iwae029. – https://spiral.imperial.ac.uk/bitstream/10044/1/113573/5/iwae029.pdf
Research brief:
This study presents a practical, card-based toolkit designed to improve nontechnical participants’ understanding of AI concepts, empowering them to actively participate in co-design sessions for AI development.
Involving the public in AI design has the potential to make AI systems more transparent, ethical, and user-friendly. Yet, the limited AI knowledge among nontechnical users often leads to misunderstandings, reducing their ability to engage in meaningful design contributions. AI technologies today often operate as ”black boxes,” with end-users unable to fully understand their workings or effects. This can lead to public misconceptions about AI’s abilities, a phenomenon sometimes called the ”Superhuman Fallacy,” where AI is assumed to have capabilities it does not possess. With transparency challenges in AI-especially related to ethical issues such as bias, privacy, and accountability-these misunderstandings can become obstacles to responsible development and informed user participation.
To bridge this knowledge gap, this research presents a practical, card-based AI literacy toolkit. It aims to introduce essential AI concepts, ethical concerns, and real-world applications to nontechnical audiences in codesign settings. Each card offers a plain-language definition, relevant example, and ”What if?” prompts that encourage critical thinking, using methods grounded in human-centred design. Tested with 50 nontechnical participants, the toolkit demonstrated significant improvements in participants’ understanding of AI topics, broadening the range of AI-related ideas and critical questions raised by over 50% and increasing relevant keyword usage by 80%. Results suggest that AI literacy tools like this one can enhance nontechnical audiences’ involvement in the AI design process, fostering a more inclusive and informed approach as AI plays an increasingly impactful role in public services and digital landscapes.
Findings:
The toolkit led to significantly improved understanding and questioning of AI concepts by nontechnical users.
Among 50 participants, those using the toolkit demonstrated a 53.94% broader range and 80.17% higher frequency of AI-related terms in their feedback compared to the control group. Terms such as ”bias,” ”dataset,” and ”model” were common among toolkit users but largely absent in the control group.
This toolkit effectively equipped participants with the vocabulary and conceptual understanding needed to engage critically in AI discussions.
Toolkit use improved collaboration and mutual understanding in group settings.
During codesign sessions, nontechnical participants reported that the toolkit ”leveled the playing field,” establishing a shared language for discussing AI ideas. Notably, it reduced discomfort by guiding discussion when gaps arose and created a common understanding of technical terms, smoothing interactions between technical and nontechnical participants.
The toolkit facilitated a shared language and mental model among participants, essential for bridging knowledge gaps in multidisciplinary teams.
The toolkit supported more diverse and imaginative ideation around AI’s applications and impacts.
Using the toolkit’s ”What if?” prompts, participants explored various outcomes, including risks and ethical considerations. For example, one participant reflected that ”more negatives than benefits” emerged as they discussed potential harms and benefits.
The toolkit effectively broadened participants’ perspectives, encouraging them to consider both positive and negative AI impacts, crucial for responsible AI design.
Nontechnical participants valued the chance to contribute their perspectives on AI design.
Participants expressed appreciation for the toolkit’s ability to facilitate their unique insights, with two expressing a desire for ongoing input to ensure their concerns, such as ethical implications, are addressed. One participant said they valued providing a ”human perspective” in a typically technical design process.
Nontechnical participants valued their inclusion, indicating a strong desire to advocate for public interests and contribute to ethical technology development.
Advice:
Practitioners need to actively incorporate more diverse perspectives in AI development, including those of non-technical and underrepresented groups. This approach challenges the biases of traditional user-centred design, ensuring that diverse perspectives are represented and fostering more equitable, participatory, and just AI systems.