Any machines performing tasks that mimic human intelligence
Machine learning and deep learning are popular subfields of AI
Processes large datasets for insights
Evolves with new data
Important Concepts
Artificial Intelligence (AI)
Machine Learning (ML)
Deep Learning (DL)
Artificial Intelligence (AI)
Broad concept of machines performing tasks in a way that mimics human intelligence.
Machine Learning (ML)
Subset of AI where machines learn from data without being explicitly programmed for specific tasks.
Deep Learning (DL)
Advanced subset of ML using neural networks with many layers to analyze data, recognize patterns, and make decisions.
Evolution of Programming
Conventional Programming
Machine Learning Programming
Deep Learning Programming
Conventional Programming
Programmer defines rules and instructions for the computer to follow.
Data is processed based these rules
Machine Learning Programming
Programmer trains model
Chooses algorithms
Provides data
Selects important features
Models
Learns patterns from data
Make predictions
Deep Learning Programming
Programmer trains model
Chooses complex neural network algorithms
Provides large dataset
Models
Learns complex patterns
Finds important features
Make predictions
What is Causing all this Excitement?
Deep Learning AI models
Specifically, Deep Learning models that can produce human-like content
These are called Generative AI models
They can generate text, images, music, and more
ChatGPT is an example
Does AI Think on Its Own?
No! people might think it does because:
Has astonishing capabilities
Limited programming effort
Mostly training on data
Unexplainable black box
AI as a Black Box
Challenges in understanding AI decision paths
Makes it unpridinctable
Efforts towards explainable AI
You can prompt the AI to explain its output
Balancing trust with transparency in AI systems
Educating users on AI capabilities and limitations
Examples of astonishing Capabilities
Recognizes complex patterns unseen by humans
Utilizes vast data to make insightful or precise predictions
Output not distinguishable from human’s
However, AI
Lacks original creativity
Relies on training data, could be biased
Lacks consciousness and self-awareness
Lack ability to recognize right from wrong
Beware of hallucinations
Requires human oversight
Key Takeaways
Generative AI produces human-like content
Recombines and generates new content from training data
Lacks consciousness and creativity
Requires human oversight
Prudence in AI to Write Research
You Have Been Using AI for Sometime in Research
Plagiarism checkers
iThenticate, Turnitin, etc.
Grammar and language checkers
Grammarly, Word, etc.
Translation tools
Google Translate, DeepL, etc.
Search engines
Google, Bing, etc.
What Changed?
AI impact is more observable with Generative AI
It can now:
Draft research papers and articles
Create and refine research questions and outlines
Simplify complex explanations
Produce and analyze data
Summarize papers and reviews
Review and edit papers
and more
ChatGPT
Latest tool to have a significant impact
What is ChatGPT?
Type of generative AI by OpenAI
It generates content
Mimics human-like text responses
Trained on diverse internet text
Based on Deep Learning
Continuously updated for relevance
Latest version is ChatGPT-4
How to Use ChatGPT
How to Use ChatGPT
Write prompts
Describing what you want
Give an example of what you want
Give document and ask for summary or refinement
Engage in conversation to refine output
Ask to refine output
Ask to explain how output is generated
Obvious Use Case by Researchers
Writing paragraphs, sections, or complete papers
O’Connor, S. & ChatGPT. (2023). Open artificial intelligence platforms in nursing education: Tools for academic progress or abuse? Nurse Education in Practice, 66, 103537. https://doi.org/10.1016/j.nepr.2022.103537
Risks
Will AI Replace researchers?
Unlikely to happen given limitations of AI
Will researchers misrepresent AI-generated content as their own?
More likely, content will be unoriginal and not reliable
Refining the output likely to take more effort than writing it from scratch
Will researchers become too reliant on AI?
Likely, risk is deskilling and loss of expertise
Can be mitigated with guidelines and training
More Productive Uses of AI in Research
Summarize articles
Explain or simplify complex concepts
Analyze or visualize data
Generate empirical data from documents
Generate synthetic data for testing models
Generate/translate research instruments
Improve wording or paragraph
Draft outlines and research questions
and more
Beware of
Ignoring AI-generated content’s need for validation
Always validate, experience in subject matter is crucial
Relying on AI to produce new content
It’s a tool, not a replacement
Write the content, ask AI to refine it
Not reporting to editors how AI was used in publication
We are still trying to understand how AI is used in research
Policies and guidelines are still evolving
Relying on first produced output
Engage in conversation to refine output
AI as a tool becomes more effective as you learn to use it
Exploring Other Tools
Elicit
Consensus
Researchrabbit
SciSpace
Litmaps
Scite
Trinka
Scholarcy
iThenticate
Key Takeaways
AI can be a powerful tool in research, that has been used for a while
New tools like ChatGPT are making AI more impactful
Other useful tools are available
AI should complement human expertise, not replace it
Guidelines on use of AI in research are still evolving
Prudence in AI in the Editorial Process
Role of editorial boards
Think about the role of AI in the editorial process
Establish guidelines and policies for AI use
Transparency will be a minimum requirement
Educate authors and reviewers on AI use
Utilizing AI in the editorial process
Assist in screening submissions
Avoid over-reliance on AI, likely to reject creative and innovative submissions
Enhance peer review process
Summarize and compare reviews
Use AI as the third reviewer
Improve communication language
Enhance feedback/reports to be polite and constructive
I would advise against using AI to write your reviews
Same reasons as using AI to write research
New Challenges
Undetectable plagiarism
AI doctored/generated data
Change in predatory practices
Possible overwhelming of conventional review process
AI generated papers passing initial screening threshold
Solutions
New tools to detect AI-generated content
Embrace AI tools and become familiar with their capabilities
Adoption of Open Science practices
Open data, open access, open peer review
Be transparent about AI use
Conclusion
AI’s potential to transform scientific publishing is immense
Responsible use and ethical considerations are paramount
Collaboration between AI and human expertise is key
Continuous evaluation of AI’s impact and evolution
Transparancy in use is necessary in moving forward
AI is a tool, not a replacement
Rather than jobs disappearing, more likely people not using it in their jobs will disappear
