AI research tools

The use of AI for text creation is not yet uniformly regulated at Trier University of Applied Sciences. Clarify with your lecturers whether and in what way AI-supported tools are permitted when writing your seminar papers or theses.

• Reliability of the results:
  Generative AI models such as ChatGPT can provide erroneous or even invented sources (“hallucinations”). Therefore, check all data critically.
   
• Scientific literacy:
  The use of AI should be understood as support. Do not miss the opportunity to develop your own research and analysis skills.
   
• Ethics & sustainability:
  The use of AI can have problematic aspects, including copyright issues and high energy consumption.
   
• Dynamic development:
  AI tools are evolving rapidly. We endeavor to keep this website up to date, but can only provide a snapshot of what is currently available.

• Niels Van Quaquebeke
  AI Tools for Academia
• Texas Tech University Libraries
  Artificial Intelligence Tools for Detection, Research and Writing
• Tübingen University Library
  Literature research with AI
• Virtual Competence Centre VK:KIWA
  AI-Tools for scientific working

Generative AI models such as ChatGPT or Copilot can support you in preparing your research. They help you to gain an initial overview of a topic, identify key search terms and formulate queries. They can also give you suggestions on where and how best to carry out your search.

In addition to this support for a thematic search, it is also possible to ask the AI for specific sources, standards or specialist literature (e.g. “Which international standard defines the requirements for environmental management systems in companies?”). Important: The content of generative AI is not verified. Models can produce so-called hallucinations, i.e. invent information or misrepresent sources. A critical examination of all proposals is therefore essential.

In order for generative AI models to provide the most helpful answers for search preparation, the quality of the query (prompt) is important.

Here are some recommendations:

• Be as specific as possible. Describe your request clearly and precisely (e.g. “Summarize the most important research questions on topic XY”).
• Provide contextual information, for example: Your subject area, aim of the research, course of study or occasion (“For my seminar paper in environmental informatics on topic XY, I am looking for ...”).
• Formulate a clear task. Ask the AI for a list, an outline, a summary or search questions, for example.
• Use examples or restrictions. Limit the time period, language or subject area (“Only German-language studies since 2020”).
• Ask for alternatives or variants. This will give you different perspectives for your research (“Give me three different approaches”).
• Do not accept the answers without checking them. Use the AI's suggestions as a starting point for further research, not as a finished result.

Example prompts:

• “I need a brief overview of current research approaches for my seminar paper in the field of environmental technology on 'Energy efficiency of heat pumps in existing buildings'. Please discuss different heat pump technologies and specifically consider the German market.”
• “Please give me a list of the most important synonyms for the term 'microplastics' and sort them by category: Origin, Impact and Removal. I need these terms for a systematic literature review on microplastics in water, focusing on publications from the last 10 years.”
• “Name three different approaches to the use of solar systems in agriculture (agri-photovoltaics). Please explain the technical, economic and ecological advantages and disadvantages of each. Make sure that you only consider studies from the last 5 years.”
• “For my project work in microbiology on the topic 'Growth inhibition of E. coli by antimicrobial substances', I need suitable search queries for the PubMed database. Please give me ten specific search queries that take into account both the bacterium Escherichia coli and factors of growth inhibition by antibiotics, plant extracts or chemical compounds.”

• OpenAI
  ChatGPT
• Microsoft
  Copilot
• DuckDuckGo
  Duck.ai
• Google DeepMind
  Gemini
• Meta
  Llama
• OpenAI
  ScholarGPT (at "Research & Analysis")

AI-supported search engines such as Semantic Scholar, Connected Papers or Elicit use artificial intelligence to find and link relevant scientific articles more efficiently.

Their special features compared to conventional search engines:

• Semantic search: they analyze the content and context of a search query and search not only for keywords, but for thematically related papers.
• Linking knowledge: They recognize content-related relationships between works and show how they are connected to each other (e.g. through citations or thematic similarities).
• Automatic summaries: Some tools create abstracts of papers to enable a quicker assessment of the content.
• Prioritization of relevant studies: AI models evaluate which abstracts are particularly important for a topic based on citations, influence and subject area.

The following tips will help you to use AI search engines effectively:

• Test out entire questions instead of individual terms (e.g. “How do microplastics affect the growth of cyanobacteria?” instead of “microplastics cyanobacteria”).
• If required, specify the desired document type (e.g. “Systematic Review”, “Meta Analysis”).
• If required, filter by year and subject area to obtain only current and relevant studies.
• Use citation networks to discover fundamental or more recent research on a topic.
• Compare the results of different search engines, as they use different algorithms.

• Consensus NLP
  Consensus - AI Search Engine for Research
• Elicit
  Elicit - The AI Research Aissistant
• Technische Informationsbibliothek TIB
  ORKG ask - Scientific Search and Discovery System
• Cactus Communications
  R Dicovery - AI tool for literature search
• scienceOS
  scienceOS - AI research agent for scientists
• Pluto Labs
  Scinapse - AI research discovery service
• Allen Institute for Artificial Intelligence
  Semantic Scholar

In addition to traditional search engines, there are specialized AI-supported tools that map and visually present the (citation) relationships between scientific works. They help to better understand the state of research on a topic by showing

• which works cite each other and are related,
• how a field of research is developing (linking older basic research works with current developments),
• whether a study is cited positively or critically, e.g. using tools with smart citations,
• where there are research gaps (thematically related works without citation links).

These tools are particularly helpful if you want to find thematically central sources or create reviews.

The following strategies will help you to use these tools effectively:

• Start with a thematically important article or work and analyze its citations.
• Look for highly cited studies to identify seminal works or particularly influential papers.
• Compare positive and critical citations to identify controversies and points of discussion (e.g. with the help of scite.ai).
• Use different tools at the same time.

• Connected Papers
  Connected Papers - Explore academic papers in a vsiual graph
• Litmaps
  LitMaps - Discover the world of Scientific Literature
• Open Knowledge Maps
  OpenKnowledge Maps - Map a research topic with AI
• Human Intelligence Technologies
  ResearchRabbit - Citation-based literature mapping tool
• Scite
  Scite - AI Research Assistant