In2writing: Accepted papers

In2writing: Accepted papers === [TOC] ## DiaryMate: Exploring the Roles of Large Language Models in Facilitating AI-mediated Journaling **Keywords:** ```! 'LLM’s generated sentences', 'Inner-state evaluation' and 'AI-based reflection' ``` **Problem:** Exploring the challenge for journalist who struggle to explain their ideas in journaling. **Proposed solution:** Presenting DiaryMate which is an LLM supporting tool for journalists to translate their ambiguous ideas and experiences into writing. The tool is highly regarded after conduct a filed study to assess it quality in journalism assist writing. Donghoon Shin, Taewan Kim, Hwajung Hong and Young-Ho Kim University of Washington, KAIST and NAVER AI Lab [ Read referenced paper](https://cdn.glitch.global/d058c114-3406-43be-8a3c-d3afff35eda2/paper23_2023.pdf) ## Writing with Generative AI: Multi-modal and Multi-dimensional Tools for Journalists **Keywords:** ```! 'Multi-modal writing', 'Generative AI systems', 'Concept Explorer' and 'Multi-dimensional writing' ``` **Discussion:** Out of a granted fund by NSF for Furthre Newes work, the authors explored ways of augmenting writing assistant for journalists. Via two complementary directions. One direction is building generative writing assistant along with image generation of storyboard. The other building network graph-based interface for exploring ideation in a semi structured yet non-linear way. LLM’s should be combined with Generative AI too. whereby other AI Tools are combined together such as text generation and image as to generate storyboards via graph based interface that allows for exploring the various dimensions before narrative ordering is arrived at. ![](https://hackmd.io/_uploads/SJQAiWFU3.png) =========== ![](https://hackmd.io/_uploads/rJv0i-FL2.png) =========== ![](https://hackmd.io/_uploads/BkB0s-FUh.png) Sitong Wang, Lydia B. Chilton and Jeffrey V. Nickerson Columbia University and Stevens Institute of Technology [ Read referenced paper](https://cdn.glitch.global/d058c114-3406-43be-8a3c-d3afff35eda2/paper25_2023.pdf) ## Can AI Support Fiction Writers Without Writing For Them? **Keywords:** ```! 'Writing support tools', 'Creativity support tools', 'Interview study', 'Fiction writing', and 'Hman-AI collaboration' ``` **Problem:** LLM’s generating generating stories like TaleBrush are posing ethical dilemma of literature ownership and artist integrity. **Proposed solution:** Authors propose providing writer via writer assistant constant updated summaries as to help story structure and assist in the writing process in line with the writer values and preferences ![](https://hackmd.io/_uploads/HkEYwWFLn.png) Jessi Stark, Daniel Wigdor, Anthony Tang, Young-Ho Kim and Joonsuk Park University of Toronto, Singapore Management University, NAVER AI Lab and University of Richmond [ Read referenced paper](https://cdn.glitch.global/d058c114-3406-43be-8a3c-d3afff35eda2/paper26_2023.pdf) ## Dimensions for Designing LLM-based Writing Support **Keywords:** ```! 'LLM capabilities', 'Task complexity', 'Output Quality' and 'Writing tasks' ``` **Problem:** Complex AI tasks requiring almost perfect AI performance delivers low quality output instead of delivering high quality outputs **Proposed solution:** To achieve an average quality in writer assistant, an experience is needed as an inspiration. Authors argue LLM experience has has three consideration: LLM Capabilities, task complexity and output quality. Proposing these consideration are complimenting common taxonomy dimensions. These dimensions are derived from human-centered and social-technical perspectives. Such as feasibility, value co-creation, usability and etc. Further arguing that taxonomy of writer assistant capturing these dimensions can create the process of designing experiences. ![](https://hackmd.io/_uploads/SyfONZtUh.png) Frederic Gmeiner and Nur Yildirim HCI Institute, Carnegie Mellon University [ Read referenced paper](https://cdn.glitch.global/d058c114-3406-43be-8a3c-d3afff35eda2/paper27_2023.pdf) ## Beyond Summarization: Designing AI Support for Real-World Expository Writing Tasks **Keywords:** ```! 'AI-Assisted Writing', 'Summarization', 'Expert Writing', 'Augmented Writing' and 'Expository Writing'. ` ``` **Problem:** The article highlights the different scenarios of AI-assisted writting except for Expository writing (Evidence Based Writing/ knowledge generating) that is currently understudied and is intended to be explored in this paper. Expository writing includes summaries of external documents as well as new information or knowledge.Expository writing can be seen as a sensemaking process. **Proposed solution:** Developing AI support tool for expository writing has fruitful research challenges. The interaction between reading and writing brings new challenges for developing AI support tool. It is important to design different types and levels of AI support to maximally help the authors while minimally influencing or shaping their opinions. Expository writing has two unique characteristics: it is evidence-driven and knowledge-generating as illustrated in Figure 1. ![](https://hackmd.io/_uploads/r17folYL3.png) **Solution challenges:** 1- LLM”s suffer from Hallucination, although strong in understanding and text generation. 2-LLM’s are limited in reasoning. 3- LLM’s are limited in long-form generation capabilities. **[Read complete summary](https://hackmd.io/dlFV0aAXR3yHwoKRXI7rXA)** ## The Model is the Message **Keywords:** ```! `Machine-generated text` `Language control ` `Language Homogenization` ``` **Analysis:** The authors explored different angles around the the new technology application of AI, LLM.Beliving the LLG as text generation medum is instigating and accelerating a cultural shifts.They further explore the language had been under abusive prower and control systematically mainly by corporate. Examples: 1- Politics of Language Control. 2- Historical Corporate Control Over Language. 3- Contemporary Corporate Control Over Language. The authors expressed their belifes to proactively explore concerns of centralized power over language production tools. Such an understanding will allow us to better design and develop tools that meet the human-centered values around language production, rather than defaulting to profit-incentivized commercial decisions. They also explored the language homogenization from different points: 1- Centralized Models and The Writing Processes. 2- Homogenization and the Writer. 3- Homogenization and the Reader. Isabelle Levent and Lila Shroff Stanford University [ Read referenced paper](https://cdn.glitch.global/d058c114-3406-43be-8a3c-d3afff35eda2/paper29_2023.pdf) ## Using Large Generative Models for Storyboarding: Challenges and Goals **Keywords:** ```! `Storyboarding` `Creative tool ` `Large generative models` ``` **Challenge description:** Producing a storyboard can be tedious and time consuming due to several factors. Firstly, writing the script itself is a challenging process, as the writer needs to transfer their high-level comprehension of the story into a structured, concrete, and coherent form. Additionally, the created visual plots has to consistently align with the written script, which requires certain expertise from the producers **Design Challenges:** 1- How to generate consistent characters and scenes in a video script. 2- How to establish the correspondence between the script and visual elements both at a specific moment and throughout the storyboard. 3- How to personalize LGMs to fit a particular creator. **Design goals:** 1-Establishing inter- and intra- modality connections in the storyboard. 2- Enabling effective human-AI interaction for intent clarifying and error handling. 3- Balancing creators’ reliance on AI and their past do- main knowledge and experience in making storyboards. Zheng Ning, , Toby Jia-Jun Li and Dingzeyu Li University of Notre Dame and Adobe Research [ Read referenced paper](https://cdn.glitch.global/d058c114-3406-43be-8a3c-d3afff35eda2/paper30_2023.pdf) ## Decoding the End-to-end Writing Trajectory in Scholarly Manuscripts **Keywords:** ```! `Human-centered computing` `Applied computing ` `Software engineering` ``` **Problem:** LLMs still struggle to provide structural and creative feedback on the document level that is crucial to academic writing **Solution:** the Authors presented a novel taxonomy that categorizes scholarly writing behaviors according to intention, writer actions, and the information types of the written data. They also provide ManuScript, an original dataset annotated with a simplified version of our taxonomy to show writer actions and the intentions behind them. ![](https://hackmd.io/_uploads/H1TqJlKI2.png) =========== ![](https://hackmd.io/_uploads/SJpcyeKLh.png) =========== ![](https://hackmd.io/_uploads/rya9JlKIh.png) Ryan Koo, Anna Martin-Boyle, Linghe Wang and Dongyeop Kang University of Minnesota [ Read referenced paper](https://cdn.glitch.global/d058c114-3406-43be-8a3c-d3afff35eda2/paper34_2023.pdf) ## Augmenting Human-AI Co-Writing with Interactive Visualization **Keywords:** ```! `Interactive visualization` `AI reasoning` `Sterotypes` ``` **Problem:** The article explore the general challenges of using LLM's to write content: 1- Mitigating ownership tensions between humans and AI. 2- Enabling human autonomy in the process. 3- Creating mechanisms for writers to understand and explore the reasoning of AI. 4- Applying NLP to complex and abstract narrative components (e.g., characterization, events, dialogue. **Prposed solution:** Hypothesize introducing a communication interface between writers and AI as an Interactive Visualization. It is a prominent method for making sense of complex AI reasoning and revealing hidden patterns from text data, to be that interface. **Limitation:** Visualization as an interface between AI and humans is not validated solution. **[Read complete summary](https://hackmd.io/p_IC3eF3SfKczITz2yITSw)** ## Towards Explainable AI Writing Assistants for Non-native English Speakers **Keywords:** ```! `Language variety` `Bais` `NNES` ``` **Problem:** Non-Native English Speakers known as NNES’s face difficulties in assessing paraphrased texts generated by AI writing assistants, largely due to the lack of explanations accompanying the suggested paraphrases. **Propsed solution:** The authors described four potential user interfaces to enhance the writing experience of NNES's using AI writing assistants. The designs focus on incorporating explanations to better support NNESs in understanding and evaluating the AI-generated paraphrasing suggestions. The design is made into four forms: 1- Confidence scores of model outputs. 2- Example sentences from credible sources. 3- Attention score. which is a comparison of usage trends of original and paraphrased suggestions. 4- Textual explanation of the suggestions. ![](https://hackmd.io/_uploads/ByxecK0u83.png) **Limitation:** The design decisions involved in presenting explanations can unintentionally introduce further bias into AI tools. For instance, the manner in which an explanation is visualized or presented to the user could influence their decision-making process. This could lead to a biased text outcome. **[Read complete summary](https://hackmd.io/wU6LvIYlSzaiiQ4oTa2k0A#)** ## Using Writing Assistants to Accelerate the Peer Review Process **Keywords:** ```! `Peer review` `AI ethics` `Quality control` ``` **Problem:** Number of papers submitted is higher than available human reviwers. leading to low-quality reviews and delayed reviews **Proposed solution:** Authors belief fully automating the peer review process is currently not feasible due to design issues. However, here are ideas as to how writing assistants can help reviewers complete review tasks: 1- Easily create a review draft: Transform reviewer bullet points into logically coherent para- graphs. Enabling reviewers to focus solely on their judgments of the manuscripts without having to spend time and energy structuring it themselves.   2- Proofreading for reviewers via Grammerly like tool. This will ensure reviews are in good quality.   3- Flag emotionally charged words or phrases, and rephrase them so acceptable and helpful feedback.  4- Judgment guidance: writing assistant can provide prompts to guide them in making notes about the paper and forming a fair judgment.   **Limitation:** the autors belive ther are technical limitations and ethical concerns. Human experts are still best placed to assess the quality of academic research papers. potential risks and rigorous regulation to ensure the confidentiality of the review process is upheld and agreed upon by all stakeholders (i.e., authors, reviewers, and editors) Shiping Chen, Duncan P. Brumby and Anna L. Cox University of Central London Interaction Centre [ Read referenced paper](https://cdn.glitch.global/d058c114-3406-43be-8a3c-d3afff35eda2/paper37_2023.pdf) ## On Pastiches, Distributions, and Appropriations **Keywords:** ```! `sociolinguistics` `discourse` `pragmatics` ``` **Discussion** Jaylen Pittman Shares his understanding lingusitic phenomenas & sociolingustic concerns when facilitating LLM's as a writing assistant in a minorized lanugage varity. Sharing his concerns that these phenonmenas can impose negative racial impact, espocially when a language model tries to write in minoritized language varity such as African American English. For some phenomnas he provides analogy of practises,implication. For others, he just explains the implication in GPT as language model. **[Read complete summary](https://hackmd.io/@rk-njMthS3alhN2pv31t9w/no_27)** ## An Engineering Perspective on Writing Assistants for Productivity and Creative Code **Keywords:** ```! 'human-AI collaboration', ' large language models' ``` **Discussion:** The article discussed in the article focuses on the use of AI/ML-based code writing assistants and their impact on the workflow of software developers. The article proposes a focus on two themes: 1-measuring the impact of writing assistants 2- how code writing assistants are changing the way engineers write code. The article discusses various metrics used to evaluate the impact of writing assistants, including behavioral and attitudinal measures. The article also highlights how code writing assistants are changing the way developers write code. The article acknowledges the challenges of using writing assistants for creative writing and the need to explore the influence of code writing assistants on creativity and novelty in the context of software engineering. Sarah D'Angelo and Ambar Murillo Google [ Read referenced paper](https://cdn.glitch.global/d058c114-3406-43be-8a3c-d3afff35eda2/paper1_2023.pdf) ## What Can’t Large Language Models Do? The Future of AI-Assisted Academic Writing **Keywords:** ```! 'large language models, AI-powered writing assistants, academic writing' ``` **Discussion:** The article discusses the potential benefits and limitations of using AI-powered writing assistants for academic writing. The authors suggest that writing assistants can help with tasks such as (literature review, writing, and revision) but caution that the accuracy and factuality of the generated text must be carefully monitored. The article also highlights the need for further research to improve the design of writing assistants and ensure that they support, rather than displace, researchers. The article discusses the potential of large language models (LLMs) in supporting academic writing processes. The authors identify five human-AI collaboration paradigms of writing assistants in creative writing: (ideation, continuation, elaboration, rewriting, and revision). They also discuss the potential of LLMs in supporting fact-checking, grammatical fixes, rephrasing text, rewriting passages for clarity, ensuring consistency of jargon, verifying logical soundness, checking for controversial or biased statements, identifying additional supporting references, and more. The authors argue that while these writing assistants may help scaffold or automate rote aspects of academic writing, they are inadequate without their co-creators. Raymond Fok University of Washington [ Read referenced paper](https://cdn.glitch.global/d058c114-3406-43be-8a3c-d3afff35eda2/paper4_2023.pdf) ## Interactive writing systems and why small(er) could be more beautiful **Keywords:** ```! 'writing assistants, data sets, large language models, cocreativity' ``` **Discussion:** The paper proposes the use of smaller and more specialized datasets, known as Small or Specialized Language Models (SLMs), as a way to mitigate the issues related to large data-driven Language Models (LLMs). The paper argues that SLMs could be more accessible, inclusive, and culturally flexible than LLMs, and could enable greater human influence over generative AI systems in creative contexts. The paper also suggests leveraging work within machine learning and bias to design writing assistants for underrepresented languages, types of writers, and writing tasks. Additionally, the paper proposes co-creative systems that cater to multiple support needs and align system capabilities with user expectations. Finally, the paper suggests using popular music, such as rap, as a source of training data for writing assistants, which could represent wider audience concerns, thoughts, and feelings and bring voices from excluded communities into the NLP community. Ibukun Olatunji swansea university [ Read referenced paper](https://cdn.glitch.global/d058c114-3406-43be-8a3c-d3afff35eda2/paper8_2023.pdf) ## Writing Assistants Should Model Social Factors of Language **Keywords:** ```! 'large language models' ``` **Discussion:** The artical discusses the factors that influence language use and variation in writing. These factors include personality, social relations and norms, time, geography, and domain language. The context emphasizes the importance of writing assistants to account for these factors to provide personalized recommendations for word choice and sentence phrasing. Examples of socio-linguistic phenomena such as honorifics, slang, code-switching, and avoidance speech are also provided. The context also highlights how meanings of words can change over time and across different dimensions. Vivek Kulkarni, Vipul Raheja Grammarly [ Read referenced paper](https://cdn.glitch.global/d058c114-3406-43be-8a3c-d3afff35eda2/paper5_2023.pdf) ## A Situation Awareness Perspective on Intelligent Writing Assistants: Augmenting Human-AI Interaction with Eye Tracking Technology **Keywords:** ```! 'writing assistant, eye tracking, situation awareness' ``` **Discussion:** The artical discusses the factors that influence language use and variation in writing. These factors include personality, social relations and norms, time, geography, and domain language. The context emphasizes the importance of writing assistants to account for these factors to provide personalized recommendations for word choice and sentence phrasing. Examples of socio-linguistic phenomena such as honorifics, slang, code-switching, and avoidance speech are also provided. The context also highlights how meanings of words can change over time and across different dimensions. Moritz Langner, Peyman Toreini, Alexander Maedche Karlsruhe Institute of Technology [ Read referenced paper](https://cdn.glitch.global/d058c114-3406-43be-8a3c-d3afff35eda2/paper9_2023.pdf) ## Creative Struggle: Arguing for the Value of Difficulty in Supporting Ownership and Self-Expression in Creative Writing **Keywords:** ```! 'intelligent writing assistant, creative writing' ``` **Discussion:** The article proposes the concept of "creative struggle" as a framework for considering the role of effort and ownership in creative writing. The authors argue that AI writing assistants should not remove all obstacles and frustrations, but rather enable writers to focus on the creative challenges that are most personally fulfilling to solve. The article discusses the importance of psychological ownership in creative writing and how AI tools can affect it. The authors suggest that the ideal writing assistant strikes a balance between enabling creative momentum and allowing sufficient creative struggle to let the writer feel ownership over the output and engage in effortful self-dialogue that results in deeper self-understanding, authentic expression, and human creativity. The article also proposes future research directions to investigate the role of creative struggle in perceptions of creative ownership and its relationship to AI assistants. David Zhou, Sarah Sterman University of Illinois Urbana-Champaign [ Read referenced paper](https://cdn.glitch.global/d058c114-3406-43be-8a3c-d3afff35eda2/paper11_2023.pdf) ## Repurposing Text-Generating AI into a Thought-Provoking Writing Tutor **Keywords:** ```! 'Text-Generating, Writing Tutor' ``` **Discussion:** The article discusses the potential of text-generating AI technology to revolutionize writing education. The authors propose repurposing AI into a thought-provoking writing tutor with the addition of recursive feedback mechanisms. They developed a prototype AI writing-support tool called Scraft that asks Socratic questions to users and encourages critical thinking. A preliminary study with 15 students in a university writing class showed that Scraft’s recursive feedback is helpful for improving their writing skills. However, participants also noted that Scraft’s feedback is sometimes factually incorrect and lacks context. The authors discuss the implications of their findings and future research directions. Tyler Taewook Kim, Quan Tan Columbia University [ Read referenced paper](https://cdn.glitch.global/d058c114-3406-43be-8a3c-d3afff35eda2/paper12_2023.pdf) ## Approach Intelligent Writing Assistants Usability with Seven Stages of Action **Keywords:** ```! 'Large Language Models, Norman's seven stages' ``` **Discussion:** The given context discusses the use of large language models (LLMs) in writing assistants and proposes a framework based on Norman's seven stages of action to design and evaluate interfaces that support fine-grained actions at different stages of interaction with LLMs,the seven stages consists of (a) goal formation, (b) plan, (c) specify, (d) perform. The framework can help identify crucial interactions and guide the design of writing assistants to support the tutorial authoring task. The context also highlights the importance of investigating human-LLM interactions to better assess the model's performance and proposes potential avenues for research in this area. Avinash Bhat, Disha Shrivastava, Jin L.C. Guo McGill University [ Read referenced paper](https://arxiv.org/pdf/2304.02822.pdf) ## Practical Challenges for Investigating Abbreviation Strategies **Keywords:** ```! 'abbreviation strategies, assistive writing, cognitive load' ``` **Discussion:** "The article discusses the challenges of investigating abbreviation strategies for assistive writing technology. The authors argue that studies on abbreviation schemes need to optimize for four goals: minimizing input time, minimizing cognitive load, successfully representing the communicative intent of the message, and accurately portraying the user's tone and stylistic details. The article presents two case studies for two distinct experimental paradigms and discusses the implications of the results. In Case Study 1, participants were given a specific abbreviation scheme to apply to auditory stimuli. The goal was to abbreviate sentences with minimal mistakes and as quickly as possible. The abbreviation scheme involved shortening each word to its initial letter and removing punctuation. Participants listened to spoken sentences of varying lengths and abbreviated the response part. The effectiveness of the scheme was established through error and reaction time analyses, and open comments. The study found that the auditory signal significantly affects how the abbreviation scheme is applied, and what constitutes a word depends on the audio signal. In Case Study 2, participants were not provided with a specific abbreviation scheme but were given a communicative goal that the message needed to achieve. Participants were instructed to abbreviate a message as much as possible while still allowing the receiver of the message to fulfill a task. Participants were allowed to delete characters and words but not add words or rewrite. In each trial, participants were given a message that needed to be shortened as much as possible, together with a communicative goal that their version of the message should still manage to achieve. The results showed that depending on the goal, users employ distinct strategies when it comes to abbreviating content, which posits a new challenge for evaluating the practical usefulness of an abbreviation expansion model." Elisa Kreiss, Subhashini Venugopalan, Shaun Kane, Meredith Ringel Morris Stanford University [ Read referenced paper](https://cdn.glitch.global/d058c114-3406-43be-8a3c-d3afff35eda2/paper17_2023.pdf) ## Parachute: Evaluating Interactive Human-LM Co-writing Systems **Keywords:** ```! 'Language models' ``` **Discussion:** The context provides information about a proposed evaluation framework called Parachute for human-LM interactive co-writing systems. Parachute identifies three key components for evaluating co-writing systems: human, LM, and writing artifact. It also recognizes three crucial aspects for evaluating the components' interactions: evaluating human-LM interaction, evaluating dynamic interaction trace, and evaluating writing artifact. The context also provides a categorized evaluation metric for each Parachute aspect. Hua Shen, Tongshuang Wu Pennsylvania State University [ Read referenced paper](https://cdn.glitch.global/d058c114-3406-43be-8a3c-d3afff35eda2/paper15_2023.pdf) ## What Writing Assistants Can Learn from Programming IDEs **Keywords:** ```! 'Integrated Development Environments,Writing Assistants ' ``` **Discussion:** "The article discusses the potential benefits of applying ideas from Integrated Development Environments (IDEs) to Writing Assistants (WAs). IDEs are software tools used by programmers to write and manage code, while WAs are AI-based tools that assist writers in various writing tasks. The article suggests that WAs could benefit from features such as code style conventions, explicit context, and structural word coloring, which are commonly used in IDEs. The article also proposes using IDEs as a model for WA research to test assumptions about interventions to help people interact with writing tools more efficiently. Integrated Development Environments (IDEs) are software applications that provide a comprehensive environment for software development. IDEs are designed to make software development more efficient and less demanding for mental effort. IDEs leverage the hallmarks of programming languages and development practices to provide assistance to programmers" Sergey Titov, Agnia Sergeyuk, Timofey Bryksin [ Read referenced paper](https://arxiv.org/pdf/2303.16175.pdf) ## Future Writing Assistants for Qualitative Research **Keywords:** ```! 'large language models, qualitative analysis ' ``` **Discussion:** "The article discusses the potential benefits of using large language models (LLMs) in qualitative analysis (QA) to reduce the time cost, offer multiple analyses, address inconsistencies associated with researcher bias, and help researchers explore themes within their data. LLMs can also be used to help with artifact generation by generating interesting questions for a given research topic. The article provides examples of model-generated thematic analyses and discusses the potential limitations of LLMs in certain qualitative research frameworks. Courtni Byun, Kevin Seppi Brigham Young University [ Read referenced paper](https://cdn.glitch.global/d058c114-3406-43be-8a3c-d3afff35eda2/paper19_2023.pdf) ## Writing Tools: Looking Back to Look Ahead **Keywords:** ```! 'writing technology, natural language processing, intelligent writing tools, interactive editing ' ``` **Discussion:** The article discusses the evolution of writing tools and the need for new tools that incorporate recent technological advancements. The article emphasizes the importance of user-friendly interfaces, support for teaching and learning how to write, easy extension of features based on user needs, and real interaction with the system that enables writers to stay in control of edits. The article also discusses the potential of AI-based components to support writers in creating and editing documents, as well as the need for structure within texts to allow for display according to features of devices and tailored to the needs of readers. Cerstin Mahlow Zurich University of Applied Sciences [ Read referenced paper](https://arxiv.org/pdf/2303.17894.pdf) ## Towards an Authorial Leverage Evaluation Framework for Expressive Benefits of Deep Generative Models in Story Writing **Keywords:** ```! 'large language models, mixed initiative, generative ai, intelligent narrative technologies, NLP, NLU, computational creativity ' ``` **Discussion:** The given text is a collection of excerpts from various sources discussing the role of AI in storytelling and comedy. It includes references to studies and case studies exploring the use of AI in generating prompts and structuring stories, as well as the potential for AI to play a comedic role. The text also includes quotes from experts in the field discussing the storytelling process and the potential for AI to automate mental tasks. Sherol Chen, Carter Morgan, David Olsen, Ethan Manilow, Mark Nelson, Qiuyi Zhang, Senjuti Dutta, Piotr W Mirowski, Kory Wallace Mathewson [ Read referenced paper](https://cdn.glitch.global/d058c114-3406-43be-8a3c-d3afff35eda2/paper22_2023.pdf) Ideas for LLMs as Writer Assistant === :::info :bulb: The following is a list of ideas subject to further Research & Implementation, some of them derived from the accepted papers above ::: - Conduct an empirical study to evaluate, visualization as an interface between AI and humans can mitigate the general challenges related to writer assistants. - Implement a a supporting tool for Fact checking related to the text generated by LLM. - Implement a supporting tool for locating coniuty problems for writers during the writing process. - Implement an experince advising support tool to bridge gaps between writer experince and what the story character is exnerincing. - Implment AI Character building assisting support tool to either: identify dialoouge and passges inconsistant to the character prior description. or provide suggestions for paraphrasing the passage. - Conduct research experimentation of DiaryMate to realize the benefit of tool without any damging to the journalist personal and emotional nature of the journalist experince. ## Tasks Performed by the System in future **Automate repetitive tasks** - Producing outputs with a standardized formatting and quality such as: AD write up, legal case characteristics, highlighting relevant law acts and regulations based on the legal cases **Accelerate writing** - Smoothing out non billable tasks editorial tasks such as, Document drafting , writing reviews, writing blogs and back & forth emails escalations of contract management **Brainstorm ideas** - Develop main legal points of the case instead of starting from scratch such as provide broader ideas for the legal write up and finding legal tip and solution **Augment a skill** - Train the module on the necessary skills of writing clear documents and contracts and explains essential legal vocabulary and phrases. **Condense information** - Obtaining references and books that briefly explain laws and regulations - Providing articles specialized in specific areas of laws and regulations - Allocating lines of communication with people who specialize in complex matters in important issues **Simplify the complex** - Understanding and analyzing the issues being asked about - Simplify legal concepts and terminology - Reformulation of legal definitions and regulations in a simpler way - Interpretation of issues and search for the best solutions **Expand perspectives** - Diversity in writing context - Generating opinions in legal issues and allowing more than one opinion or legal option - Add arguments and evidence to support the cases **Improve what’s available** - Focus on correcting spelling errors - Pay attention to the connection of ideas - The possibility of printing the result of definitions, legal arguments and proofs on paper and ease of use by the user