Bianca's Digital Garden


Search IconIcon to open search

AI-Enabled Design Futuring

Last updated May 6, 2024

# Overview

# Inspiration

# Ideas

# Notes

# 7/17/23

# Project Outline

# Draft 1

# Project Overview

Title of Project: Keywords: Generative AI, Design Futuring, Speculative Design,

# Background and Current Approaches

# Rationale and Objectives

# Research Questions

  1. In what ways does AI support the generation of ideas for an utopian future for youth in developing countries?
  2. What are utopian and dystopian futures typically imagined by youth in developing countries, and what role does technology play in these?
  3. How might design fictions as a method of futuring elicit new ideas that speak to concepts of a utopian reality for youth in developing countries?


# Hypothesis Statements

# Conditions and Variables

# Possible Conditions

# Possible Variables

# Technical Script

to see how were gonna do the studies/workshops.

# Targeted Contributions

# Evaluations

# Study 1: Pilot Study Protocol
# Study Design
# Measure (Analysis Criteria)
# Results

# Study 2: Testing with Focus Group

# Study 3 (Complementary): Workshop Setting

# To-Do

# Resources

# Draft 2

# Sources

# For describing the problem
# Possible design guidelines
# Prompt engineering
# Example projects
# **Explaining Cards/Design Futuring

# Writing

# Outlines
# Introduction
# Background/Related Work
# Methodology
# Final
# Introduction

The future of technological innovation in the Global South rests partly on the next generation of researchers. These early career researchers (ECRs) are motivated by the possibility of contributing to scientific development and to wider society, along with the desire to be novel or innovative. Unfortunately, many of them struggle in designing and undertaking this kind of impactful research due to a lack of funding, support, visibility and recognition. Additionally, in the human-computer interaction (HCI) field, ECRs could also be blocked by the limitations of established methods.

One such example is human-centered design (HCD): “an approach to interactive systems development that aims to make systems usable and useful by focusing on the users, their needs and requirements, and by applying human factors/ergonomics, and usability knowledge and techniques”. Its widespread influence can be found in its core adoption across many computing, design, and management projects in industry, academia, and government. But while it is effective at delivering popular and effective interactive systems, HCD’s focus on individual people or groups gives it a narrow temporal and contextual focus.

One limitation of HCD is its limited perception of time, which can be seen in designers’ tendency to design something new for the current world, instead of designing a completely new world. This is because HCI design and development is an iterative engineering process, wherein design is gradually improved by a growing understanding of the users and the context of use. Thus, while HCD may be suitable for incremental improvements, it is a not a fit driver of breakthrough innovation. Another flaw of HCD is that it often misses out on ethical, societal, and political aspects. HCD and its ISO, ISO 9241-210, have been criticized for lacking consideration of user security and privacy, along with being too rigid to adapt to local cultural contexts. And since human values are often not touched on in HCD, HCI researchers usually do not critically evaluate the potential consequences of their work.

If ECRs in the Global South believe that high-quality research must be beneficial and groundbreaking, then it is crucial for them to utilize more radical design methods. Enter design futuring: a type of approach to design that is “concerned with future alternatives” and seeks to “produce knowledge through [discourse]” (instead of being used to solve an immediate problem). It not only helps design researchers generate and communicate new ideas, but also contemplate the social and technical implications of their work. However, producing design fictions often requires proficiency in creative software, preexisting assets, and time — things many researchers may lack, especially if they’re located in the Global South. Thus, there is a need for the development of creative tools that can support rapid prototyping for design futuring.

Given its meteoric development, generative AI (i.e. AI technologies that automatically generate written or visual content based on text prompts) has the potential to bring these kind of tools to life. In line with this, we propose an AI-enabled creativity support tool that rapidly envisions future aspects of a researcher’s area of interest, building on Ivanov et. al’s design futuring sprint method. This paper is a preliminary study that explores how generative AI can help ECRs in the Global South imagine, evaluate, and consider the implications of their works, so that they can work towards high-quality innovation. This study aims to make a dataset of prompts for fine-tuning a model designed to generate fictional depictions of the future; this would then be used in the development of the proposed creativity support tool.

Background: Design Futuring in HCI Research + Practice

Related Work: AI-Enabled Creative Tools for Design Futuring Generative AI tools have made creative exploration more accessible, thanks to rapid advancements in large language models (LLMs) and diffusion models. Popular examples of these tools, such as ChatGPT and Dall-E, implement a low-floor, high-ceiling and wide-walls principle: they have a low barrier of entry (low floor), allowing users to achieve high-quality outputs (high-ceiling) that can extend to extend to a broad range of applications (wide walls). Acting as an “engine for the imagination”, generative AI has been specifically been used to support the creative processes of ideation and creation. While it has already been applied to creative activities like visual storytelling and culinary innovation, it is most especially prominent in design, with fields like industrial design, fashion design, and architecture.

Despite generative AI’s widespread adoption in several creative activities, its use in speculative and critical design remains underexplored. But generative AI has already shown to be useful across a spectrum of envisioning speculative futures, from practical forecasting to creative storytelling. Applications of generative AI have started to demonstrate their potential for collective design futuring, encouraging discourse on societal issues like climate change. Overall, all of these examples show that “AI can serve as a subjective interface linking fiction and reality”. Our project contributes to this growing body of work by exploring how generative AI can be act as an interface that connects the realities of researchers with the futures they desire.

There is an opportunity to assess how generative AI can play a more active role — what if it was not just a tool, but also a collaborator? AI-driven co-creative tools are less involved in the early stages of co-creation (i.e. researching, incubating). This is because of the misconception that machines lack the ability to evaluate and distinguish mundane outputs from masterpieces, along with concerns for accountability, responsibility, and fairness in human-AI interaction. However, the data-driven, non-human mind of AI offers much potential for redefining and repositioning early stages of the creative process; even without setting creativity as the target for optimization, AI can introduce intriguing perspectives that may fall out of humans’ sights. For example: when Lyu, Hao, and Yi embedded generative AI into the design futuring process, it served multiple functions: signal scanning, scenario development, world-building, and more.

But AI must still be used with caution; without intentional human guidance, machine-generated visions may be harmful and insubstantial. For instance, in response to Seattle leaders using AI imagery to promote a “downtown of the future”, Wolfe discussed the following challenges that AI-enabled design futuring can bring about: (1) the data being used to train AI may contain biases and inequalities, which may lead AI to perpetuate marginalization; (2) AI may be unable to account for evolving needs, given that it is trained on a limited set of data; (3) without definitional prompts, AI lacks contextual understanding, which can lead to generic outputs; and (4) AI may prioritize standards like efficiency and aesthetics over the human experience. To ensure human-centered outputs, an AI-enabled design futuring tool will require meticulous prompt engineering. It will also require user education. Usually, users are forced to be hands-off when using AI co-creative tools due to a lack of knowledge on how the tools work, what level of complexity to expect in the output, and how machines optimize their behaviors. This particularly occurs with generative AI; when users get unsatisfying outputs, they have limited guidance for making improvement. In light of these considerations, this study would like to investigate how to develop a thoughtful AI futurist.

# Methodology

We will be conducting a pilot study primarily aimed at answering our research question: How can generative AI best support the design futuring process?

Participants Through convenience sampling, we will be recruiting five participants: computer science students working at a university research laboratory in the Philippines. Given their technical background, they are expected to already be familiar with using generative AI tools (e.g. ChatGPT, Dall-E). They should also be currently working on a research project, since they will be required to explore it in the pilot study.

Study Design The participants will go through a 90-120 minute workshop based on the first day of Ivanov et. al’s design futuring sprint. We will be using a within-subjects design where all participants do the activities on their own, and then try doing the same activities with the help of generative AI. The workshop will consist of the following parts:

Documentation and Analysis The workshop will take place remotely via Zoom, allowing for easy recording. We will be using a shared Figjam whiteboard throughout the whole workshop; all prompts and outputs (i.e. answers, images) will be placed here. To evaluate the participants’ experience with generative AI, we will be transcribing and conducting a thematic analysis of the FGD.

# Limitations and Future Work