SCIGEN 101/101G Notes: Generative AI in Academia (Waipapa Taumata Rau)

Course Context and Objectives

  • Waipapa Taumata Rau, University of Auckland; Course code SCIGEN 101 / 101G
  • Topic: Can AI support quality communication?
  • Date: August 8, 2025August\ 8,\ 2025

Notices

  • Milestone should now be in (Monday)
  • Assignment #1 due August 27August\ 27
  • Join Manaaki Study Buddy Groups
  • Learning essentials – resources for students

Lecture Objectives

  • Gain awareness of the different ways AI is being used in academic research and science communication
  • Explore and understand the ethical considerations and other safeguards currently used/discussed in the academic community
  • Understand the broader context of AI and quality science, quality communication and trust in scientific evidence

What is Generative AI?

"Generative AI is a type of artificial intelligence that creates new content, such as text, images, audio, and video, based on what it has learned from existing data. Unlike traditional AI that analyzes or categorizes data, generative AI produces original content by learning patterns from large datasets and using algorithms to mimic human creativity"

How should we work with Gen-AI in undergraduate teaching and learning?

"If 1st year students are using Gen-AI for assignments, penalties should be seriously considered. If 4th year students are NOT using Gen-AI for assignments, penalties should be seriously considered." - Professor in Social Sciences at a large comprehensive university in Canada, on the launch of ‘Anthropology of the Internet’

What do you think?

  • What was this prof thinking about?
  • What are the benefits and risks of using Gen-AI in universities?

Potential Benefits for Students (from UoA Guidelines for Students)

  • Help you to revise content (e.g. generating practice questions or flashcards) ext(practicequestions,flashcards)ext{(practice questions, flashcards)}
  • Get an overview of your topic before you begin exploring in detail
  • Help you to improve your writing skills, including grammar, use of vocabulary, and sentence structure
  • Explain solutions to problems in another way to help you understand how a particular answer was arrived at (e.g. in mathematics)
  • Help you to analyse data and/or organise information

Potential Risks (from UoA Guidelines for Students)

  • Accuracy: Always fact check
    • Non-factual
    • Inaccurate
    • Out-of-date
  • Bias: Always be critical
    • Reflect biases found within the data sources used to teach gen AI
    • Discrimination of marginalised groups
    • Under-representation of marginalised groups
  • Privacy: Be cautious
    • Information and data uploaded will be remembered potentially used
    • Personal details cannot be protected once uploaded
    • Confidential information or data can be widely disseminated
  • Quality: Be discerning
    • may lack originality
    • may not have the tone that you are trying to achieve for your assessment, or be in appropriate style for your purpose
    • May be repetitive or poorly structured
  • Note: Co-Pilot from UoA remains in a protected space within the university

Can Gen-AI help me review the literature?

  • Limitations for literature discovery in literature reviews:
    • Quality
    • Depth
    • Scope
  • Methodological considerations:
    • Comprehensiveness
    • Biases
    • Reliability (both input and output)
    • Transparency
    • Reproducibility
  • AI tools for literature reviews (Woods, 2025)

Cases and Cautions

  • When you could use Gen-AI:
    • When the required rigour is lower
    • Personal knowledge building
    • Keeping up to date
  • How you should use Gen-AI:
    • Fact check and challenge the outputs
    • Provide the AI a lot of context in prompts
    • Iterate through prompts
    • Record and declare use

Let’s have a go…

How did we use Co-Pilot?

  • Become familiar with the topic?
  • Sense check our understanding?
  • Quickly obtain key sources? Check these!
  • Generate keywords for a rigorous database search?
  • Iterate and challenge. Was there saturation?
  • Upload and analyse data? … No
  • Write an essay? … No

What Is GenAl?

  • GenAl is a type of narrow AI tool trained to conduct a specific task (i.e., respond to text-based prompts)
  • What Are Human Responsibilities When Using GenAI?
    • Transparent disclosure and attribution
    • Verification and oversight of content and analyses
    • Documentation of generated data
    • Focus on ethics and equity
    • Continuous monitoring, oversight, and public engagement

DO:

  • Use tools to support efficiency
  • Disclose use
  • Develop human-machine collaborative workflows
  • Keep track of changing best practices

DO NOT:

  • List as author
  • Use for figure manipulation or generation
  • Use to conduct peer reviews
  • Human: Analyze data
  • Create tables

Example Workflow

  • GenAl
    • Provide preliminary sense-check on data
    • Interpret
    • Come up for key points
    • Develop text sections
    • Cite appropriate literature
    • Discuss methods or findings
    • Review grammar and logical flow
    • Provide feedback on missing details

Considerations

  • How will GenAl use change communication, knowledge, and advancement of science?
  • GenAl challenges us to question what makes humans special

The Scholarly Communication Cycle

  • A slow and steady conversation – RELIABILITY of KNOWLEDGE
    • Peer review
    • Replicability
    • Building on the knowledge record
    • Contributing to the knowledge record
  • Research → Knowledge Creation
  • Manuscript → Peer Review Publication (e.g. Academic Journals)
  • Dissemination: Print, online, libraries, web …
  • Preservation (e.g. archiving in literature databases)
  • Building on knowledge: cited, recombined, applied

Systemic Considerations

  • Integrity of Science: not just individual projects but broader system implications of building on AI-generated outputs
  • Ethics and governance: Access to literature and data is still uneven globally
  • FAIR principles (Findable, Accessible, Interoperable, Reusable)
  • CARE principles (Collective Benefit, Authority to Control, Responsibility, Ethics)
  • National science systems must develop robust data governance policies in the first instance as Gen-AI moves at pace
  • International Science Council Webinar and case studies

Gen-Al to Improve Science Communication (3 models)

  • Think about the 3 communication models in applying Gen-Al to science communication

Three Models of Communication (Grisé, 2012)

  • 1) Linear Model: Transmission of message moved from sender to receiver
  • 2) Interactive Model: Message moved from sender to receiver – feedback sent from receiver back to the original sender
  • 3) Transactional Model: Dynamic process of communication involving ‘transactions’ between the participants leading to the co-creation of meaning

Useful Resources for Students

  • Module from UoA: video, self-assessment, resource material
  • Deakin University: Guide to AI literacy
  • Sydney University: module for AI literacy, with advice on prompts and use cases

AI Usage Standards for Waipapa Taumata Rau

  • The data classification of any Inputs submitted to GenAI tools must be established
  • The choice of GenAI tools must be restricted to those suitable for the data classification level of inputs submitted:
    • Public Data: Any appropriate GenAI tool may be used
    • Internal and Sensitive Data: Must only be used where a negotiated contract and service agreement exists between the University and the GenAI provider that establishes adequate protection for Inputs (adequate protection ensures Inputs are not used for any other purpose by the provider, including further training of their public GenAI)
    • Restricted Data: Only services solely controlled by the University may be used
  • A Privacy Impact Assessment must be completed before a GenAI tool is used with Personal Information
  • The designated owner of a business function within the University is accountable and responsible for validating GenAI output prior to use of that output to inform business processes within their remit
  • Users of GenAI tools should consult with the Office of the Pro-Vice Chancellor Māori where Māori data may be used in a GenAI tool, or use may impact Māori
  • Users of GenAI tools should familiarise themselves with the limitations and/or the possibility of inherent bias within the tool prior to use
  • Any content (including text, image, or video) intended for publishing or distribution where a substantial portion of the content has been created by a GenAI tool should be labelled as such

Prompt Example

  • Create a photo realistic illustration of a scientist reacting to the results of a generative AI prompt
  • Question: Why does the scientist appear so shocked?

Commentary on the Illustration Prompt (Page 21-22)

  • The scientist appears shocked because the results could be unexpected, groundbreaking, or unsettling
    • Possible interpretations include: unexpected accuracy, ethical implications, scientific breakthrough, humorous or surreal outputs, existential realization
  • Model bias explanation: The image depicted a male scientist by default due to the generative model’s default bias; training data overrepresents certain demographics in scientific roles
  • These notes illustrate how bias can manifest in AI-generated visuals and the importance of deliberate prompting and awareness of training data biases

Practical Implications for Exams and Practice

  • Understand how AI tools can support learning while recognizing limitations and safeguards
  • Be able to articulate potential benefits and risks in research design, literature reviews, writing, and data analysis
  • Apply the Scholarly Communication Cycle as a framework for analyzing AI-assisted scholarly work
  • Discuss ethical, governance, and equity considerations in adopting GenAI in academic settings
  • Differentiate between linear, interactive, and transactional communication models and apply them to science communication scenarios