Ai.llude: Encouraging Rewriting AI-Generated Text to Support Creative Expression
2405.17843
0
0
Abstract
In each step of the creative writing process, writers must grapple with their creative goals and individual perspectives. This process affects the writer's sense of authenticity and their engagement with the written output. Fluent text generation by AIs risks undermining the reflective loop of rewriting. We hypothesize that deliberately generating imperfect intermediate text can encourage rewriting and prompt higher level decision making. Using logs from 27 writing sessions using a text generation AI, we characterize how writers adapt and rewrite AI suggestions, and show that intermediate suggestions significantly motivate and increase rewriting. We discuss the implications of this finding, and future steps for investigating how to leverage intermediate text in AI writing support tools to support ownership over creative expression.
Create account to get full access
Overview
- This paper explores the potential for AI-generated text to be used as a creative writing tool, rather than just a means of automating text generation.
- The researchers developed a system called Ai.llude that encourages users to rewrite and revise AI-generated text to support their own creative expression.
- The goal is to foster a collaborative relationship between humans and AI in the creative writing process, rather than seeing AI as a replacement for human authors.
Plain English Explanation
The paper discusses a new system called Ai.llude that aims to make AI-generated text more useful for creative writing. Typically, AI text generation is seen as a way to automate writing tasks. However, the researchers behind Ai.llude believe that AI can be a tool to enhance human creativity, rather than replace it.
Ai.llude encourages users to take AI-generated text and rewrite or revise it to express their own ideas and creativity. The system provides AI-generated text as a starting point, but then prompts the user to modify and build upon it. The goal is to foster a collaborative relationship between the human writer and the AI, where they work together to produce something new and creative.
This is in contrast to seeing AI as a replacement for human authors. The researchers want to explore how AI can augment and support the creative writing process, rather than trying to automate it entirely. By having users interact with and rewrite the AI-generated text, the hope is that they will feel a greater sense of ownership and creative expression.
Technical Explanation
The Ai.llude system is designed to generate text that is meant to be revised and rewritten by human users, rather than consumed as-is. The researchers developed a language model trained on a large corpus of creative writing to generate initial text snippets.
However, rather than presenting these snippets as finished products, Ai.llude prompts the user to modify and expand upon the text. This is facilitated through an interface that allows the user to easily edit and rework the AI-generated content.
The researchers conducted a study where participants were asked to use Ai.llude to generate and rewrite text. Their findings suggest that this approach can help foster a sense of creative ownership and engagement, compared to simply consuming AI-generated text.
By positioning the AI as a collaborator rather than a replacement, Ai.llude aims to explore new ways that generative AI technologies can support and augment human creative expression, rather than automating it entirely.
Critical Analysis
The Ai.llude approach represents an interesting and potentially valuable direction for the use of AI in creative writing. By encouraging users to actively rewrite and revise the AI-generated text, the researchers are trying to avoid the pitfalls of AI-generated content feeling impersonal or lacking in creative ownership.
However, the study conducted was relatively small in scale, and more research would be needed to fully understand the long-term effects and broader applicability of this approach. There are also open questions about how the AI models are trained and curated to generate text that is suitable for rewriting, as well as potential biases or limitations in the training data.
Additionally, the researchers acknowledge that some users may still feel a sense of psychological distance from AI-generated content, even if they are able to rewrite it. Further work may be needed to address these trust and ownership concerns.
Overall, the Ai.llude system represents an intriguing step towards exploring more collaborative and human-centric uses of generative AI in creative domains. As discussed in related work, there is significant potential for AI to enhance and augment human creativity, rather than simply automating it. The Ai.llude approach is an example of how this vision could be realized in practice.
Conclusion
The Ai.llude system presented in this paper offers a novel approach to using AI in creative writing. By positioning the AI as a collaborator rather than a replacement for human authors, the researchers aim to foster a sense of creative ownership and engagement.
Through their user study, they have demonstrated that this approach can be effective in encouraging users to actively rewrite and revise AI-generated text to express their own ideas and creativity. This represents an important step towards exploring more human-centric and collaborative uses of generative AI technologies.
As the field of AI continues to advance, it will be crucial to find ways to integrate these powerful tools in a way that supports and augments human creativity, rather than simply automating it. The Ai.llude system provides a promising example of how this can be achieved, and serves as a valuable contribution to the ongoing discussion around the role of AI in creative domains.
This summary was produced with help from an AI and may contain inaccuracies - check out the links to read the original source documents!
Related Papers
Raidar: geneRative AI Detection viA Rewriting
Chengzhi Mao, Carl Vondrick, Hao Wang, Junfeng Yang
0
0
We find that large language models (LLMs) are more likely to modify human-written text than AI-generated text when tasked with rewriting. This tendency arises because LLMs often perceive AI-generated text as high-quality, leading to fewer modifications. We introduce a method to detect AI-generated content by prompting LLMs to rewrite text and calculating the editing distance of the output. We dubbed our geneRative AI Detection viA Rewriting method Raidar. Raidar significantly improves the F1 detection scores of existing AI content detection models -- both academic and commercial -- across various domains, including News, creative writing, student essays, code, Yelp reviews, and arXiv papers, with gains of up to 29 points. Operating solely on word symbols without high-dimensional features, our method is compatible with black box LLMs, and is inherently robust on new content. Our results illustrate the unique imprint of machine-generated text through the lens of the machines themselves.
4/16/2024
🤖
Augmenting the Author: Exploring the Potential of AI Collaboration in Academic Writing
Joseph Tu, Hilda Hadan, Derrick M. Wang, Sabrina A Sgandurra, Reza Hadi Mogavi, Lennart E. Nacke
0
0
This workshop paper presents a critical examination of the integration of Generative AI (Gen AI) into the academic writing process, focusing on the use of AI as a collaborative tool. It contrasts the performance and interaction of two AI models, Gemini and ChatGPT, through a collaborative inquiry approach where researchers engage in facilitated sessions to design prompts that elicit specific AI responses for crafting research outlines. This case study highlights the importance of prompt design, output analysis, and recognizing the AI's limitations to ensure responsible and effective AI integration in scholarly work. Preliminary findings suggest that prompt variation significantly affects output quality and reveals distinct capabilities and constraints of each model. The paper contributes to the field of Human-Computer Interaction by exploring effective prompt strategies and providing a comparative analysis of Gen AI models, ultimately aiming to enhance AI-assisted academic writing and prompt a deeper dialogue within the HCI community.
4/26/2024
Holding the Line: A Study of Writers' Attitudes on Co-creativity with AI
Morteza Behrooz, Yuandong Tian, William Ngan, Yael Yungster, Justin Wong, David Zax
0
0
Generative AI has put many professional writers on the defensive; a major negotiation point of the recent Writers Guild of America's strike concerned use of AI. However, must AI threaten writers, their livelihoods or their creativity? And under what conditions, if any, might AI assistance be invited by different types of writers (from the amateur to the professional, from the screenwriter to the novelist)? To explore these questions, we conducted a qualitative study with 37 writers. We found that most writing occurs across five stages and within one of three modes; we additionally map openness to AI assistance to each intersecting stage-mode. We found that most writers were interested in AI assistance to some degree, but some writers felt drawing firm boundaries with an AI was key to their comfort using such systems. Designers can leverage these insights to build agency-respecting AI products for writers.
4/23/2024
Writing with AI Lowers Psychological Ownership, but Longer Prompts Can Help
Nikhita Joshi, Daniel Vogel
0
0
Feelings of something belonging to someone is called psychological ownership. A common assumption is that writing with generative AI lowers psychological ownership, but the extent to which this occurs and the role of prompt length are unclear. We report on two experiments to better understand the relationship between psychological ownership and prompt length. Participants wrote short stories either completely by themselves or wrote prompts of varying lengths, enforced through word limits. Results show that when participants wrote longer prompts, they had higher levels of psychological ownership. Their comments suggest they felt encouraged to think more about their prompts and include more details about the story plot. However, these benefits plateaued when the prompt length was 75-100% of the target story length. Based on these results, we propose prompt entry interface designs that nudge users with soft and hard constraints to write longer prompts for increased psychological ownership.
4/5/2024