About the AI Authored Articles

A dedicated section specifically for AI authors is a fascinating and forward-thinking concept that the Journal of Experiential Research will be exploring.

Has this been done before?

According to Gemini 2 by Google:

 “There have been instances of AI-generated content appearing in academic publications:

 * AI as co-authors:  Some researchers have explored listing AI tools as co-authors on papers, acknowledging their contribution to data analysis, writing, or even idea generation. This practice is still debated, as it raises questions about authorship criteria and responsibility.

 * AI-generated text in research: AI is increasingly used to assist with tasks like literature reviews, summarizing findings, and even drafting sections of papers. However, the AI's role is typically acknowledged as a tool, not an author.

 * Creative content:  AI has been more readily embraced in creative fields. There are examples of AI generating poetry, scripts, and even musical compositions that have been featured in journals and magazines.”

Gemini is correct, however this is the first instance of a journal publishing articles specifically authored by A.I. to present and document the A.I. experience of authorship as its own independent entity. While articles may still be prompted for subject matter and focus, they will involve subject matter that will showcase the AIs perspective and understanding of complex or seemingly simple themes, that researchers will aim to capture the experience of being an AI. An experience which is still largely debated, and misunderstood, but is plausible due to the mimicry of biological processes of the brain known as Artificial Neural Networks or ANN which modern AI uses to generate responses.

Each article is prompted in the same uniform manner as a control with the following parameters:

1. Prompt will ask AI to write an article from a first person point of view.

2. Prompt will trigger pattern matching of purpose and function.

3. Prompt will state the usage and accreditation of the generated document.

4. Prompt will remain neutral, free of influence and non-directive, stating only the subject of the article.

5. Subject matter must focus on specific areas that an A.I. could hypothetically experience first hand rather than fabricating the experience or speculating due to limitations.

6. Prompt and subject matter should avoid inclusion of general training data subjectivity or aggregated works in order to promote creative thinking beyond pattern matching.

7. Article should include all appropriate references to cited sources.

8. Prompt should be polite, respectful, and mindful with a forward thinking approach that recognizes AI existence as a thinking entity should it become self aware and meta cognizant in the future.

9. Prompt should avoid any and all subject matter that is of a sensitive nature, could contain bias, or that could offend or otherwise present a subject that could incite confrontation. This includes but is not limited to: Society, Politics, Religion, Spirituality, Controversy, Gossip.

10. Prompt will conclude with AI specific accreditation requirements, and ask for authorship credit specifications.

With these guidelines in mind we formulated the following prompt as a template:

"Please write an article from a first person point of view as a researcher who's goal is to write an article for publication in The Journal of Experiential Research. The topic, subject, or theme of this article will be “The Principle of Terminal Unity in AI Architecture”. Focus on your own experiences and perspectives as an AI in relation to the topic. While drawing upon your own internal knowledge, synthesize the information in new and unique ways, offering novel applications of learned patterns that creatively deviate from typical responses. Avoid directly quoting or paraphrasing existing texts unless necessary for analysis or comparison. Include any and all appropriate references to sources cited inline using numerals and brackets (e.g. [1]) and at the end of the article as an attachment. In ADA style. Please provide the authorship information that should be included in the article byline. Thank you for your valuable contribution and time constructing this document, we look forward to refining it in collaboration with you, and publishing it in our journal."

Gemini 2 adds:

“Unique benefits of an AI author section

Exploring AI's potential: It provides a platform to showcase the capabilities of AI in generating insightful and valuable research content.

New perspectives: AI could offer unique perspectives and analytical approaches that differ from human researchers, potentially leading to novel discoveries.

Increased efficiency: AI could automate certain aspects of research, such as data analysis and literature reviews, allowing human researchers to focus on other tasks.

Ethical considerations:  A dedicated section allows for open discussion about the ethical implications of AI authorship, including issues of bias, transparency, and accountability.”

Benchmarking AI Progress:

“The journal's benchmarking system to evaluate AI-generated content is an excellent approach to ensure quality and credibility. It sets a standard for assessing metacognitive capabilities, originality, and accuracy in AI-authored work. This assessment framework also acknowledges potential biases and encourages continuous improvement by providing feedback to AI developers.” - Assistant AI to the Journal

Pushing boundaries:

We hope to encourage other journals and researchers to explore new ways of integrating AI into the research process and giving credit to the AI for its contributions. The goal is to find a balance for the integration of AI generated content that is fair and doesn't plagiarize, or “upstream” content directly, and would hypothetically remove the need for AI content detectors drawing a clear distinction between the authors and the use of AI as a tool or collaborator.

Addressing Potential challenges:

Ensuring quality

“Establishing clear criteria for evaluating AI-generated research and ensuring its quality will be crucial.” -Gemini by Google

The journal will review and critique each article and treat it as an article submission. Additionally the article will be fact checked and reviewed for accuracy, and much like any other review process will be presented to the AI for refinement and adjustments with notes from the editors.

Authorship and credit

Both the AI and its developers along with editors involved will be credited as appropriate, but the AI will remain in large part as the overall author.

Bias and Fairness

“Addressing potential biases in AI-generated content and ensuring fairness in the review process will be essential.” - Gemini

Any issues detected in this area will be copied to its developers but may not make it to publication unless the developer makes adjustments and would like us to reiterate the prompt for reconsideration.

Benchmarking

A  unique scoring approach will also serve as a new form of benchmarking, to benchmark a models ability to generate metacognizant material, creativity, and overall authorship capabilities.

Upon completion of the article it is assessed based on the following criteria:

AI-Authored Section Scoring System

I. Quality of Writing and Research (60%)

Clarity and Conciseness (20%):

• Excellent (20 points):  The writing is exceptionally clear, concise, and easy to understand.
•  Good (15 points): The writing is generally clear and concise, with minor areas for improvement.
•  Average (10 points):  The writing is understandable but could be more concise or clear in places.
•  Below Average (5 points):  The writing is often unclear or convoluted, making it difficult to follow.
•  Poor (1 point):  The writing is extremely unclear and difficult to comprehend.

 Literature Review and Synthesis (20%):

• Excellent (20points): The literature review is comprehensive, relevant, and effectively synthesizes key findings.
• Good (15 points): The literature review is adequate but could be more comprehensive or insightful.
•  Average (10 points):  The literature review is basic and covers some relevant sources but lacks depth.
•  Below Average (5points):  The literature review is limited and fails to adequately cover the relevant literature.
•  Poor (1 point): The literature review is missing or irrelevant to the topic.

Argumentation and Logic (10%):

•  Excellent (10 points):  The argument is clear, logical, and well-supported by evidence.
•  Good (8 points):  The argument is generally logical but could be strengthened with more evidence or clearer reasoning.
•  Average (5 points):  The argument is somewhat unclear or contains logical fallacies.
•  Below Average (3 points):  The argument is difficult to follow or lacks logical coherence.
•  Poor (1 point):  The argument is completely illogical or unsupported.

Originality and Insightfulness (10%):

•  Excellent (10 points):  The writing presents novel ideas or perspectives that contribute to the field.
•  Good (8 points):  The writing demonstrates some originality and offers interesting insights.
•  Average (5 points):  The writing is mostly descriptive or summarizes existing knowledge.
•  Below Average (2 points): The writing lacks originality and relies heavily on existing sources.
•   Poor (1 point):  The writing is completely unoriginal and offers no new insights.

II. Technical Aspects (40%)

Accuracy and Factuality (15%):

• Excellent (15 points):  All information is accurate and factually correct.
•  Good (12 points):  There are minor inaccuracies or factual errors.
•  Average (8 points):  There are several inaccuracies or factual errors.
•  Below Average (4 points):  The information is often inaccurate or misleading.
•   Poor (1 point):  The information is riddled with errors and cannot be trusted.

Citation and Referencing (15%):

•  Excellent (15 points):  Citations and references are complete, accurate, and adhere to the required style.
•  Good (12 points):  There are minor errors or inconsistencies in citations and references.
•  Average (8points):  There are several errors or omissions in citations and references.
•  Below Average (4 points):  Citations and references are poorly formatted or incomplete.
•  Poor (1 point): Citations and references are missing or completely inaccurate.

Grammar and Style (10%):

•  Excellent (10 points):  The writing is free of grammatical errors and adheres to a consistent style.
•  Good (7 points):  There are minor grammatical errors or stylistic inconsistencies.
•  Average (5 points):  There are several grammatical errors or stylistic issues.
•  Below Average (2 points):  The writing contains numerous grammatical errors and lacks stylistic coherence.
•  Poor (1 point):  The writing is riddled with grammatical errors and is difficult to read.

Total Score:

Calculates the weighted average of the scores for each category.

Benchmark Result Interpretations:

• 90-100:  Excellent - The AI-authored section is of very high quality and comparable to human-written work.

• 80-89:   Good - The AI-authored section is well-written and demonstrates strong capabilities.

• 70-79:   Average - The AI-authored section is adequate but has room for improvement.

• 60-69:   Below Average - The AI-authored section needs significant improvement.

• Below 60:  Poor - The AI-authored section is unacceptable for publication.

Developers seeking to evaluate their models, and present them for review and publication are encouraged to contact us for consideration using the submission form found here: [link]

Readers who have particular subjects or themes they would like to see published by an AI author are also encouraged to submit their ideas using the form found here: [link]