Enhancing Academic Writing with AI

00:53:34
https://www.youtube.com/watch?v=H8FPUiIdbwo

الملخص

TLDRThis video presentation, led by experts Jessica Parker and Kimberly Becker, delves into the use of artificial intelligence (AI) to improve academic writing by addressing both productivity and quality. The session focuses on how AI can assist in overcoming common writing challenges such as idea synthesis, structuring, and maintaining authorial identity, without compromising ethical standards. Key tools discussed include Moxy, an AI feedback tool, and Lit Maps, which aids in research. The experts explain strategies for enriching writing depth, stance, and engagement markers, which are crucial in maintaining human-like qualities in writing. Additionally, they highlight ethical issues like simplification bias, data privacy, and proper AI acknowledgment in the writing process. The presentation encourages a balanced use of AI to enhance rather than replace human expertise and emphasizes learning and adapting these tools with an awareness of their limitations.

الوجبات الجاهزة

  • 🎓 AI can enhance academic writing by aiding in synthesis and structure.
  • 🤖 Moxy provides AI-based feedback for early career researchers.
  • 📝 Lit Maps visualizes research connections for better understanding.
  • 🔍 Simplification bias is a key ethical concern with AI summaries.
  • 📚 Use AI to improve writing depth and maintain human engagement.
  • 🛡️ Data privacy is crucial when using AI tools in research.
  • ⚖️ Balancing quality and speed is essential when integrating AI.
  • ✍️ AI can help overcome writer's block by brainstorming ideas.
  • 🗣️ Maintaining authorial identity is important in AI-assisted writing.
  • 🔑 Contextual understanding is necessary for effective AI use.

الجدول الزمني

  • 00:00:00 - 00:05:00

    تەنتەشتە ئۆتكەرشى ئۈچۈن AI نى قانداق قوللاندىشى ھەققىدە كۆپ پىكىر-مۇنازىرە قىلىندى.

  • 00:05:00 - 00:10:00

    AI ۋە ھەقىقى باشقۇرۇش ئارقىلىق ئافغان كىلاسسگا ئاخىرقى كىلىشىدىن بۇرۇن باشقا ڧىلوسوفىيلارنى تەييارلىق قىلىش.

  • 00:10:00 - 00:15:00

    AI ئۈستىدە ئىلمىي تەتقىقاتنىڭ تەرتىپ پىلانى، ئۇنىڭدىكى قىستۇرۇشچا سۈپەت ۋە ۋاقىت سادہلاشمۇ ھۆججەتلەشتۈرۈليلدىغان بىلىنمايدۇ.

  • 00:15:00 - 00:20:00

    AI غا تەۋەككەلچىلىك، ئەكىلىك كەچۈرۈلمەسلىك ھەققىدە مەخپىيەتلىكتىكى مەشغۇلاتتىن ئىشلىتىلگەن.

  • 00:20:00 - 00:53:34

    AI ۋە ئىشچىلارنى قانداق قوللانسا بولىدۇ؟ كەلگۈسىدە ئۇسلۇبنى قانداق ئۇزارتسا بولىدۇ؟

اعرض المزيد

الخريطة الذهنية

فيديو أسئلة وأجوبة

  • Who are Jessica Parker and Kimberly Becker?

    Jessica Parker and Kimberly Becker are experts in artificial intelligence and academic writing. Jessica Parker is the co-founder and CEO of Moxy, a generative AI company, and a doctoral supervisor. Kimberly Becker specializes in applied linguistics and technology, and co-founded Moxy.

  • How can AI assist in academic writing?

    AI can help with writing challenges such as synthesis of ideas, structure organization, overcoming writer's block, developing authorial identity, and engaging the audience effectively.

  • What is simplification bias in AI?

    Simplification bias refers to AI's tendency to oversimplify complex concepts, leading to incomplete or inaccurate understanding, especially noticeable in AI-generated summaries.

  • What ethical considerations are there when using AI in writing?

    Ethical considerations include data privacy, avoiding oversimplification bias, ensuring quality over speed, and acknowledging AI's role in the writing process when necessary.

  • Which AI tools were discussed in the session for academic writing?

    Tools like Moxy and Lit Maps help users enhance productivity by assisting with idea synthesis, research, and providing structured feedback.

  • What first steps were recommended for those new to using AI in writing?

    The experts suggest using AI to brainstorm and generate ideas if you face writer's block, or receive feedback on your work to improve quality.

  • What are the differences between AI-generated and human-written academic text?

    AI-generated text can often lack depth, personal insights, and have mechanical precision, making it appear less nuanced compared to human writing.

  • How can AI be used to improve the quality of academic writing?

    By asking AI to evaluate your text for aspects like engagement, stance, and reporting verbs to ensure the writing maintains a human touch and depth.

عرض المزيد من ملخصات الفيديو

احصل على وصول فوري إلى ملخصات فيديو YouTube المجانية المدعومة بالذكاء الاصطناعي!
الترجمات
en
التمرير التلقائي:
  • 00:00:00
    welcome everyone tonight we are here or
  • 00:00:02
    tonight for us good morning for you to
  • 00:00:05
    talk about enhancing academic writing
  • 00:00:07
    with AI artificial intelligence and
  • 00:00:10
    we're really going to focus on balancing
  • 00:00:13
    how to ethically produce your
  • 00:00:14
    productivity while aiming for Quality a
  • 00:00:18
    little bit about us so I'm Dr Jessica
  • 00:00:20
    Parker I'm the co-founder and CEO of
  • 00:00:23
    Moxy we are a generative AI company we
  • 00:00:26
    build formative feedback tools for early
  • 00:00:30
    career researchers and graduate students
  • 00:00:32
    I also have two academic Consulting
  • 00:00:34
    companies where we work with PhD
  • 00:00:37
    students across the US and I'm also a
  • 00:00:40
    doctoral supervisor at the Massachusetts
  • 00:00:43
    College of Pharmacy and Health Sciences
  • 00:00:45
    which is in Boston Massachusetts I
  • 00:00:48
    supervise um doctoral students who have
  • 00:00:50
    professional degrees and are now getting
  • 00:00:52
    a terminal degree so I work across
  • 00:00:54
    disciplines with pharmacists physical
  • 00:00:57
    therapists occupational therapists
  • 00:00:58
    acupuncturists practitioners you may
  • 00:01:01
    know and everything we talk about today
  • 00:01:03
    are are things that we're actually doing
  • 00:01:05
    in practice either at Moxy through our
  • 00:01:07
    research or through my work with my do
  • 00:01:11
    World students Kimberly you want to
  • 00:01:12
    introduce yourself sure hi everybody I'm
  • 00:01:15
    Kimberly Becker and um I'm in Ames Iowa
  • 00:01:19
    Jessica and I are in two different parts
  • 00:01:21
    of the country in the US and yeah I'm a
  • 00:01:24
    co-founder of Moxy and my background is
  • 00:01:27
    in Applied Linguistics and technology so
  • 00:01:29
    so I have taught at the community
  • 00:01:32
    college high school community college
  • 00:01:34
    and university level I study
  • 00:01:36
    disciplinary academic writing in Corpus
  • 00:01:38
    Linguistics and usually the applied part
  • 00:01:41
    of my Linguistics is that I am a writing
  • 00:01:45
    coach and I help people with their
  • 00:01:48
    specifically with their research writing
  • 00:01:50
    so not just academic writing but
  • 00:01:52
    specific to English for research
  • 00:01:54
    publication purposes so this looks like
  • 00:01:56
    a simple agenda but we have a lot to
  • 00:01:58
    talk about today we're going to start by
  • 00:02:01
    discussing this idea of productivity
  • 00:02:03
    when we think about Ai and academic
  • 00:02:05
    writing we're then going to talk about
  • 00:02:08
    very specific techniques that you can
  • 00:02:10
    use to improve the quality of your
  • 00:02:12
    academic writing using AI notice that
  • 00:02:14
    I'm not talking about speed talking
  • 00:02:16
    about quality and then we're going to
  • 00:02:18
    talk about various ethical
  • 00:02:20
    considerations when using AI in your
  • 00:02:22
    scholarly work we'd love to kick it off
  • 00:02:25
    in the Q&A in the chat feel free to
  • 00:02:28
    answer one of these questions
  • 00:02:30
    what AI tools have you tried for
  • 00:02:32
    academic writing or what's your biggest
  • 00:02:35
    concern about using AI in your work or
  • 00:02:39
    what would make this session most
  • 00:02:41
    valuable for you if you could just take
  • 00:02:43
    a moment to enter your responses in the
  • 00:02:46
    chat that would be really great and I'm
  • 00:02:49
    going to have Kimberly keep an eye on
  • 00:02:51
    that and when we get to a natural break
  • 00:02:52
    in the discussion we're going to address
  • 00:02:55
    some of your responses we'll start with
  • 00:02:57
    writing productivity with AI I want to
  • 00:02:59
    start start by just acknowledging a lot
  • 00:03:01
    of the writing challenges that really
  • 00:03:04
    anyone faces who is conducting research
  • 00:03:07
    either for the first time or the 100th
  • 00:03:09
    time I continue to face a lot of these
  • 00:03:11
    challenges synthesis is a big one coming
  • 00:03:14
    up with new ideas based on the ideas of
  • 00:03:17
    other Scholars and combining multiple
  • 00:03:19
    sources writer's block is something that
  • 00:03:22
    is very common argumentation this really
  • 00:03:24
    working against this idea that academic
  • 00:03:27
    writing is neutral or objective
  • 00:03:29
    structure certain organization that
  • 00:03:30
    could mean outlining that could be
  • 00:03:32
    really zooming into the paragraph level
  • 00:03:34
    or even the sentence level authorial
  • 00:03:37
    identity this is really how you identify
  • 00:03:40
    as a scholar and as a researcher
  • 00:03:43
    oftentimes as novice graduate students
  • 00:03:45
    or PhD students because you're not
  • 00:03:47
    experts yet you might really question
  • 00:03:49
    your voice and whether you should have
  • 00:03:51
    an opinion or whether you're getting it
  • 00:03:53
    right and that's authorial identity and
  • 00:03:56
    then audience engagement we're always
  • 00:03:58
    writing to a specific audience and these
  • 00:04:00
    are just really common challenges there
  • 00:04:02
    are just a few that we believe AI can
  • 00:04:05
    help with in an ethical way where you're
  • 00:04:07
    not violating academic Integrity Norms
  • 00:04:10
    so when we talk about academic writing
  • 00:04:12
    with AI we are looking at it in terms of
  • 00:04:15
    a spectrum it's not a binary decision do
  • 00:04:18
    you use it or do you not use it we have
  • 00:04:21
    on one end of the spectrum what we all
  • 00:04:23
    know to be true which is human writing
  • 00:04:25
    and then we have these sort of new three
  • 00:04:28
    new areas where you have human in the
  • 00:04:30
    loop that's a lot of the writing that we
  • 00:04:32
    do as researchers is very human in the
  • 00:04:34
    loop driven where the human is making
  • 00:04:36
    the decisions we're driving the car so
  • 00:04:39
    to speak in terms of how we're going to
  • 00:04:41
    use the machine or AI there's machine in
  • 00:04:44
    the loop writing this is often more
  • 00:04:46
    appropriate for like low stakes
  • 00:04:48
    activities I know Kimberly and I used
  • 00:04:50
    the example of sometimes we'll do
  • 00:04:52
    machine in the loop writing when it's
  • 00:04:54
    like a LinkedIn post or an email because
  • 00:04:56
    that's something that's low stakes and
  • 00:04:58
    then fully synthetic writing that's
  • 00:05:00
    fully AI generated writing and depending
  • 00:05:04
    on the type of task you're engaging in
  • 00:05:07
    you might fall at different ends of this
  • 00:05:10
    spectrum here's an example of human in
  • 00:05:12
    the loop writing this came out of my
  • 00:05:14
    work with my do students I conducted a
  • 00:05:17
    participatory research study uh last
  • 00:05:20
    fall with my doc students who were
  • 00:05:22
    learning academic and scholarly writing
  • 00:05:24
    and I was particularly interested in
  • 00:05:27
    really integrating AI slowly through
  • 00:05:29
    throughout the course and seeing what
  • 00:05:31
    that meant for the new writing process
  • 00:05:34
    that they were engaging in so I allowed
  • 00:05:36
    them to engage with generative AI I did
  • 00:05:38
    give them very specific boundaries and
  • 00:05:40
    how they could use it but otherwise I
  • 00:05:43
    really just engaged them and had them
  • 00:05:45
    meet with me periodically to give me
  • 00:05:47
    feedback I had them reflect and document
  • 00:05:50
    their process and this is really what
  • 00:05:52
    came through is this idea of you're
  • 00:05:54
    collaborating with an AI and so we all
  • 00:05:57
    know that writing is a very iterative
  • 00:05:59
    process
  • 00:06:00
    we typically start with ideating we're
  • 00:06:02
    brainstorming and then we can go to
  • 00:06:04
    drafting we might go back to ideating
  • 00:06:06
    but then what's interesting is now we
  • 00:06:08
    have this opportunity to use AI for
  • 00:06:10
    maybe reviewing acting as a pure to
  • 00:06:13
    provides feedback on drafts that's like
  • 00:06:15
    formative feedback evaluation AI can
  • 00:06:18
    play a role in that as well as long as
  • 00:06:20
    you're critically thinking about the
  • 00:06:21
    output and then there's the typical
  • 00:06:23
    revision process and then sometimes you
  • 00:06:26
    go back to ideating so we know that this
  • 00:06:28
    is a complicated hled web it's not a
  • 00:06:31
    simple Circle or even a linear process
  • 00:06:34
    but in the center here I think this is
  • 00:06:35
    what's really interesting when we think
  • 00:06:37
    about human in the loop writing and this
  • 00:06:38
    is part of what I've learned from
  • 00:06:40
    Kimberly as a linguist is this idea of
  • 00:06:43
    meaning negotiation when you're
  • 00:06:45
    interacting with an AI and in this case
  • 00:06:47
    I'm specifically talking about like an
  • 00:06:49
    AI text generator an AI chatbot we're
  • 00:06:52
    continually negotiating meaning so the
  • 00:06:55
    real gold the gem of that process is not
  • 00:06:58
    just asking question and getting a
  • 00:07:00
    response it's asking for clarification
  • 00:07:03
    asking for explanations correcting it
  • 00:07:06
    even and so as long as you're
  • 00:07:07
    negotiating that meaning throughout the
  • 00:07:10
    process then you're going to get the
  • 00:07:12
    most out of that interaction and so this
  • 00:07:14
    is an ex just one example of what human
  • 00:07:16
    in the loop writing might look like when
  • 00:07:18
    you're engaging with AI when we think
  • 00:07:21
    about productivity there are and this
  • 00:07:23
    isn't even the full stack but there are
  • 00:07:25
    so many AI research assistants coming
  • 00:07:27
    onto the market I hear about new new
  • 00:07:29
    ones every day we're constantly adding
  • 00:07:31
    new logos to this Slide the one that I
  • 00:07:34
    love is lip mat so I'm going to use that
  • 00:07:36
    as an example but as you can see there
  • 00:07:38
    are many and these AI research
  • 00:07:40
    assistants all have different features
  • 00:07:42
    they're all powered by different LL lims
  • 00:07:45
    and they have a different Corpus and
  • 00:07:47
    they all do very similar but also
  • 00:07:49
    slightly different things and in terms
  • 00:07:51
    of productivity I think about this in
  • 00:07:53
    terms of working smarter not harder so
  • 00:07:56
    it's not necessarily speeding up the
  • 00:07:59
    literature
  • 00:08:00
    Discovery process it's just another tool
  • 00:08:02
    in your toolbox to help you find
  • 00:08:05
    literature that's relevant to your
  • 00:08:07
    research topic so we have a little table
  • 00:08:09
    here I'm not going to go into detail
  • 00:08:11
    with this but a lot of people like this
  • 00:08:13
    table so we're including it in the
  • 00:08:15
    presentation which you'll get a copy of
  • 00:08:17
    um these are just us comparing a few of
  • 00:08:19
    the most popular AI research assistants
  • 00:08:22
    in terms of their key features and the
  • 00:08:24
    Data Corpus that use and keep in mind
  • 00:08:26
    that the Data Corpus is evolving so you
  • 00:08:28
    could think of the Corpus is just the
  • 00:08:30
    database what literature they're pulling
  • 00:08:32
    from whenever you ask it to give you
  • 00:08:34
    search results based on whatever your
  • 00:08:36
    research topic is or your keywords are a
  • 00:08:39
    lot of these use semantic scholar which
  • 00:08:41
    I'm going to talk about in a moment and
  • 00:08:42
    a limitation of semantic scholar but
  • 00:08:44
    feel free to look at that whenever we
  • 00:08:46
    give you a copy of this presentation
  • 00:08:49
    let's just look at one of these this is
  • 00:08:50
    my favorite which is called lit Maps uh
  • 00:08:53
    and that's because I'm a visual learner
  • 00:08:55
    and so that's what I mean a lot of these
  • 00:08:56
    tools do the same thing that they have
  • 00:08:58
    different features that appeal to
  • 00:09:00
    different preferences that you might
  • 00:09:02
    have so with lit match you go in and you
  • 00:09:04
    can ask a question you can put in your
  • 00:09:06
    keyword you can just put in a few
  • 00:09:08
    Concepts and what it's immediately going
  • 00:09:10
    to do is produce a map for you and their
  • 00:09:13
    whole value proposition is that they
  • 00:09:15
    help you find what they call a seed
  • 00:09:17
    article you might also think of this as
  • 00:09:19
    a seminal piece of research where a lot
  • 00:09:22
    of research F emerg from that article
  • 00:09:25
    and what they do in this interface is on
  • 00:09:27
    the left hand side is they give you a
  • 00:09:29
    sum
  • 00:09:30
    and then on the right hand side along
  • 00:09:32
    this X and Y AIS all those little
  • 00:09:34
    bubbles represent a source and when you
  • 00:09:37
    look at the bottom we see how current
  • 00:09:39
    the further you go to the right is the
  • 00:09:42
    more current it is and on the other side
  • 00:09:44
    the further you go up is how well-sited
  • 00:09:47
    it is and so visually this helps me
  • 00:09:50
    significantly as a researcher instead of
  • 00:09:52
    just scrolling through lots of search
  • 00:09:54
    results lit Maps does have a paid
  • 00:09:57
    version of their product but like many
  • 00:09:59
    of these companies they give you access
  • 00:10:01
    to it for free so that you can try it
  • 00:10:04
    out I highly recommend checking out
  • 00:10:06
    different AI research assistants before
  • 00:10:08
    you commit to one or go all in on any
  • 00:10:10
    sort of annual subscription but this is
  • 00:10:13
    just one example of how you could
  • 00:10:16
    leverage these AI research assistants in
  • 00:10:18
    terms of productivity and just making
  • 00:10:21
    sure that your research is higher
  • 00:10:23
    quality that you're finding everything
  • 00:10:25
    that's relevant to your topic one key
  • 00:10:28
    way that the these tools differ from say
  • 00:10:31
    a traditional semantic search like
  • 00:10:33
    Google Scholar or your library database
  • 00:10:36
    or sorry keyword search so typically
  • 00:10:38
    when you go into your library or
  • 00:10:39
    databases it uses your keywords and
  • 00:10:41
    it'll scan and it look for we'll look
  • 00:10:43
    for a match of those keywords in the
  • 00:10:45
    title and the abstract and that's why
  • 00:10:47
    part of your search strategy involves
  • 00:10:49
    identifying synonyms or using wild cards
  • 00:10:52
    or truncation AI is able to scan the
  • 00:10:55
    entire article and automatically do that
  • 00:10:57
    so it makes it a little easier to find
  • 00:10:59
    literature that is relevant to your
  • 00:11:02
    topic so often I'll find articles that I
  • 00:11:04
    didn't find during my normal search
  • 00:11:06
    process so we're definitely not
  • 00:11:08
    advocating for replacing it I'll show
  • 00:11:11
    you a graph in a moment that talks about
  • 00:11:13
    triangulation but let's really think
  • 00:11:15
    about the database so semantic scholar
  • 00:11:18
    is a database that's used by many of
  • 00:11:20
    these AI research assistants it's the
  • 00:11:22
    primary one and that's because it's open
  • 00:11:24
    source it's open access and when you
  • 00:11:27
    look at the literature that's included
  • 00:11:29
    within their database you can see here
  • 00:11:31
    and these are in the millions most of
  • 00:11:34
    them 60 million are uncategorized so we
  • 00:11:37
    don't even know what discipline they
  • 00:11:38
    represent and then the next category as
  • 00:11:40
    you can see is medicine and then biology
  • 00:11:43
    and physics and you can just see it go
  • 00:11:44
    down again these are in the million so I
  • 00:11:47
    pointed to Linguistics because that's
  • 00:11:49
    Kimberly's discipline 1.1 million why do
  • 00:11:52
    you need to know this if it's not
  • 00:11:54
    obvious to you I really want to point
  • 00:11:56
    out what the limitation would be if you
  • 00:11:58
    were not looking at the database that's
  • 00:12:01
    powering a tool you might inadvertently
  • 00:12:03
    be using a tool that's not built for
  • 00:12:05
    your discipline and so it's important to
  • 00:12:08
    do your homework and look at the
  • 00:12:10
    database that's being used to power
  • 00:12:12
    these AI research assistants oops that
  • 00:12:14
    clicked on the study and so oftentimes
  • 00:12:18
    we think about a gap in the literature
  • 00:12:20
    but with these AI research assistants a
  • 00:12:22
    gap could be in the Corpus not just the
  • 00:12:25
    literature and this came to me because I
  • 00:12:27
    had a student in the humanity who was
  • 00:12:29
    using site aai which is predominantly
  • 00:12:31
    for biomedical research and they came to
  • 00:12:34
    me and they were like there's nothing on
  • 00:12:36
    this topic I was like L because you're
  • 00:12:37
    using site and it doesn't have
  • 00:12:39
    literature in the humanities so that's
  • 00:12:41
    not a gap in the literature that's a gap
  • 00:12:43
    in the Corpus and there is a difference
  • 00:12:45
    and it's really important to point that
  • 00:12:47
    out because we're definitely being sold
  • 00:12:49
    on this idea of productivity and
  • 00:12:51
    efficiency and optimization and a lot of
  • 00:12:53
    times these limitations of these tools
  • 00:12:56
    are not being advertised for us you have
  • 00:12:58
    to dig for this information and it's
  • 00:13:00
    really important again to understand the
  • 00:13:02
    difference between a gap in the Corpus
  • 00:13:04
    and a gap in the literature so when we
  • 00:13:06
    think about how we use these tools I
  • 00:13:08
    have this triangle because I like to use
  • 00:13:09
    this idea of triangulation I still use
  • 00:13:12
    Google Scholar as I normally would
  • 00:13:14
    Library databases as I normally would
  • 00:13:16
    and then I just add on AI research
  • 00:13:18
    assistance normally I add that in last
  • 00:13:21
    and I'll just end up finding additional
  • 00:13:23
    sources that I didn't find through my
  • 00:13:25
    normal search process and so this idea
  • 00:13:27
    of triangulation is really critical all
  • 00:13:30
    right I'm going to turn it over to
  • 00:13:31
    Kimberly and she's going to talk about
  • 00:13:33
    some very specific techniques that you
  • 00:13:35
    can use hi everybody yeah so just to
  • 00:13:39
    start us off if you can advance the
  • 00:13:41
    slide I just want to see what your ideas
  • 00:13:45
    are about what is a strong marker of
  • 00:13:48
    human written academic writing because
  • 00:13:51
    as we have more AI writing coming to us
  • 00:13:55
    either from students or possibly even in
  • 00:13:58
    Publications I think it's really
  • 00:14:00
    important that we know what strong human
  • 00:14:02
    writing looks like so that we can start
  • 00:14:04
    to differentiate we are not Advocates of
  • 00:14:07
    AI detectors at all they are quite
  • 00:14:10
    inaccurate we don't have anything about
  • 00:14:11
    that in this particular webinar but we
  • 00:14:15
    have a big we have a lot of webinars on
  • 00:14:17
    YouTube and if you want to know more we
  • 00:14:19
    have a whole series on AI detection and
  • 00:14:22
    you can find out more about that there
  • 00:14:25
    but yeah I'm just curious when you are
  • 00:14:28
    maybe grading student writing if you're
  • 00:14:30
    a professor or if you are a graduate
  • 00:14:33
    student what kinds of things are you
  • 00:14:35
    seeking to do in your writing and I am
  • 00:14:39
    gonna give I'm just gonna keep on going
  • 00:14:42
    after you I'll keep an eye on the chat
  • 00:14:45
    okay thanks yeah actually while they're
  • 00:14:48
    talking I did want to say Jessica in
  • 00:14:51
    case you don't have time to read all the
  • 00:14:53
    most of them are very new to Ai and
  • 00:14:56
    they're not using very many AI tools yet
  • 00:14:59
    they're a blank slate so to speak many
  • 00:15:02
    of them yeah so this diagram shows one
  • 00:15:07
    of many Frameworks for understanding
  • 00:15:09
    writing and it actually comes from a
  • 00:15:11
    study of second language writing which
  • 00:15:13
    is my field but I use it because as an
  • 00:15:16
    academic writing coach and a researcher
  • 00:15:18
    I believe that academic communication is
  • 00:15:22
    like a second dialect nobody really
  • 00:15:24
    speaks it as a quote unquote home or
  • 00:15:27
    native language and so it's relevant as
  • 00:15:31
    we are either learning how to do it
  • 00:15:33
    ourselves or teaching students how to do
  • 00:15:35
    it and so the framework has three
  • 00:15:39
    aspects and they vary in terms of their
  • 00:15:43
    importance and the very top of the
  • 00:15:45
    inverted triangle is depth depth has to
  • 00:15:48
    do with the variety and L and range of
  • 00:15:51
    language ideas sources and Analysis that
  • 00:15:54
    you are using so depth is labeled as
  • 00:15:57
    number one because this is where you
  • 00:15:58
    always want to start with your writing
  • 00:16:01
    and it's what gives your writing that
  • 00:16:04
    really human element I'll show you in a
  • 00:16:07
    minute but what we know is that AI is
  • 00:16:10
    pretty good at flow and pretty good at
  • 00:16:13
    accuracy and so depth is important and
  • 00:16:16
    it's the largest area of the triangle
  • 00:16:18
    because of that and because it really
  • 00:16:20
    allows us to get that sophistication and
  • 00:16:23
    complexity that needs to be in
  • 00:16:25
    scientific writing flow is the second
  • 00:16:28
    element and that has to do with the
  • 00:16:30
    smooth expression clear connections and
  • 00:16:32
    fluency of communication so how the
  • 00:16:35
    ideas are connected and accuracy is the
  • 00:16:39
    final kind of tip of the triangle which
  • 00:16:42
    has to do with precise mechanics
  • 00:16:44
    spelling grammar punctuation and
  • 00:16:45
    citation format if you've been thinking
  • 00:16:49
    that precise mechanics spelling grammar
  • 00:16:51
    punctuation and citation format is what
  • 00:16:53
    makes you a weaker writer I'm sorry but
  • 00:16:57
    you're wrong those are very
  • 00:16:59
    surface level issues and technology is
  • 00:17:02
    actually really good at helping us in
  • 00:17:03
    that domain and so the order that we
  • 00:17:06
    usually present when I'm coaching I talk
  • 00:17:09
    about is save the accuracy piece for
  • 00:17:11
    last and so because of that today we're
  • 00:17:14
    going to talk really about depth and I
  • 00:17:16
    want to show you some aspects of depth
  • 00:17:18
    that I think will help you so there's
  • 00:17:21
    some new research that's come out a
  • 00:17:22
    couple of new studies exploring
  • 00:17:24
    differences in text generated by
  • 00:17:26
    machines or AI versus humans of course
  • 00:17:29
    I'm very interested in this these
  • 00:17:31
    studies compared linguistic features and
  • 00:17:34
    they're the basis of much of what we're
  • 00:17:36
    going to talk about here in this next
  • 00:17:38
    section these are topics that you can
  • 00:17:40
    see a lot of in our YouTube channel but
  • 00:17:43
    I want to go over these because I think
  • 00:17:45
    it's really critical that we start to
  • 00:17:48
    know what to look for as the world gets
  • 00:17:52
    populated with more and more machine or
  • 00:17:55
    synthetic writing so the markers of
  • 00:17:58
    generated text are here there's four
  • 00:18:00
    primary ones lack of depth is number one
  • 00:18:03
    and again that goes back to that
  • 00:18:04
    inverted triangle so AI can produce text
  • 00:18:08
    that is coherent and grammatically
  • 00:18:09
    correct but it typically lacks the depth
  • 00:18:11
    and Nuance necessary for original
  • 00:18:13
    arguments or complex interpretation I
  • 00:18:15
    noticed that many of you wrote something
  • 00:18:17
    about argument in your your answer here
  • 00:18:21
    on the chat and you're definitely on to
  • 00:18:23
    something that's where the complexity
  • 00:18:25
    comes in mechanical Perfection or
  • 00:18:28
    sometimes I call this
  • 00:18:29
    synthetic accuracy or Precision this is
  • 00:18:32
    when and you've probably seen this if
  • 00:18:34
    you used an AI before it outputs this
  • 00:18:36
    thing that is very it looks okay on the
  • 00:18:39
    surface right it's grammatically perfect
  • 00:18:42
    there's really nothing that you notice
  • 00:18:44
    immediately maybe some repetition maybe
  • 00:18:47
    some words like Del everybody is picking
  • 00:18:50
    on the poor word Del it's getting a bad
  • 00:18:53
    reputation but AI basically produces
  • 00:18:56
    sentences that on the surface look okay
  • 00:18:59
    and that's what we mean there repetition
  • 00:19:01
    I've already said that U redundancy lots
  • 00:19:04
    of repeated words and if you give it
  • 00:19:06
    something in the prompt it's going to
  • 00:19:07
    pick up on that and then um duplicate
  • 00:19:10
    those words throughout its own
  • 00:19:12
    generation of text and then the last one
  • 00:19:15
    is just this absence of personal insight
  • 00:19:17
    and emotion of course AI is lacking
  • 00:19:20
    completely in personal experiences it
  • 00:19:22
    has no feelings and so it's missing that
  • 00:19:25
    Personal Touch or emotional depth even
  • 00:19:27
    in academic writing we have some
  • 00:19:30
    emotional rhetoric involved in the
  • 00:19:32
    language so let's go ahead and dig a
  • 00:19:35
    little bit deeper now into some of these
  • 00:19:37
    features and talk about what constitutes
  • 00:19:39
    death a lot of my research has been
  • 00:19:42
    related to this idea that academic
  • 00:19:44
    writing is not objective we think of
  • 00:19:47
    science as this objective faceless thing
  • 00:19:50
    but it it's not at all it's really
  • 00:19:53
    argumentative because it's acknowledging
  • 00:19:55
    and constructing and negotiating all
  • 00:19:57
    these social Rel ations of the other
  • 00:20:00
    researchers it is a conversation between
  • 00:20:03
    Scholars and so to do that we use stance
  • 00:20:06
    features we use evaluative features and
  • 00:20:09
    we acknowledge alternative views and so
  • 00:20:12
    with that in mind I did a webinar a few
  • 00:20:15
    weeks back on and and the title of this
  • 00:20:18
    is a little misleading how to assess
  • 00:20:20
    writing without AI detectors but what it
  • 00:20:22
    does is it goes into depth on all of
  • 00:20:25
    these things I'm going to tell you it
  • 00:20:26
    Dives even deeper so if you're
  • 00:20:28
    interested in any of this I'm going to
  • 00:20:30
    scratch the surface but you can dive
  • 00:20:32
    deeper through this webinar let's start
  • 00:20:35
    with stans's positionality of an author
  • 00:20:38
    to a topic amidst all the other voices
  • 00:20:40
    and by Voices the sources that they are
  • 00:20:42
    citing so this is where we evaluate the
  • 00:20:45
    subject matter we assess the status of
  • 00:20:47
    knowledge and we manifest our position
  • 00:20:50
    or stance through certainty and doubt so
  • 00:20:53
    if we're taking a very neutral position
  • 00:20:56
    the reader is questioning okay who's
  • 00:20:58
    side are you on here which direction are
  • 00:21:01
    you driving the car and so stances are
  • 00:21:04
    really important and what we know about
  • 00:21:05
    novice versus expert academic writers is
  • 00:21:08
    that novices tend to do things like they
  • 00:21:10
    boost they boost their language a lot so
  • 00:21:13
    they don't add a lot of hedging and
  • 00:21:16
    doubt and that's what I'm talking about
  • 00:21:18
    when I say their novice writers tend to
  • 00:21:19
    be very certain and if you've ever
  • 00:21:21
    taught like freshmen undergrads or in
  • 00:21:24
    high school like at 12th graders you
  • 00:21:26
    know the highest level um of secondary
  • 00:21:28
    school they're very certain about
  • 00:21:30
    everything and some of that is
  • 00:21:31
    developmental but a lot of it is just
  • 00:21:33
    not really understanding how to use
  • 00:21:36
    doubt or hedge their writing and be
  • 00:21:39
    careful about their claims so I'm going
  • 00:21:41
    to walk you through now um some AI
  • 00:21:43
    generated writing and I'm going to show
  • 00:21:45
    you how to inject some of some stance
  • 00:21:49
    and some other um depth features into
  • 00:21:52
    your writing let's imagine that you're
  • 00:21:54
    using a very popular AI writing tool
  • 00:21:57
    that's called quillbot
  • 00:21:59
    um quillbot looks something like this
  • 00:22:01
    it's actually free you can type a little
  • 00:22:03
    bit in it and then a little pop up over
  • 00:22:06
    on the side as you can see this little
  • 00:22:08
    this little yes that lightning bolt you
  • 00:22:10
    can click that and this popup comes up
  • 00:22:12
    and it says generate ideas complete this
  • 00:22:15
    paragraph start a new paragraph add an
  • 00:22:17
    example Etc and so I did this with some
  • 00:22:21
    writing I continued or just finished a
  • 00:22:24
    paragraph there and if you'll go to the
  • 00:22:26
    next slide I'll show you what I put in
  • 00:22:29
    now this is some writing that we had
  • 00:22:31
    this first top paragraph there is fully
  • 00:22:34
    human this is our um framework for AI
  • 00:22:38
    literacy in higher ed Jessica and I have
  • 00:22:41
    published a white paper on this and so
  • 00:22:43
    you can see that it's it's related to
  • 00:22:45
    what are we going to do the debate the
  • 00:22:48
    controversy about the rapid adoption of
  • 00:22:50
    AI and higher ed and basically this
  • 00:22:53
    issue of Educators and students learning
  • 00:22:55
    to understand it and at that point where
  • 00:22:57
    you see the ellipse Mark after the word
  • 00:22:59
    Technologies I clicked finish the
  • 00:23:01
    paragraph and this is what quilot
  • 00:23:04
    generated so I've bolded words that show
  • 00:23:09
    repetition basically Educators and
  • 00:23:12
    students it's repeating higher ed it's
  • 00:23:14
    repeating gaps in understanding it's
  • 00:23:17
    repeating the word leverage and some of
  • 00:23:19
    you said repetition not duplicative
  • 00:23:22
    language and that sort of thing and
  • 00:23:24
    that's for sure what's happening here
  • 00:23:26
    and this is very common in AI writing
  • 00:23:30
    and so now go go ahead to the next slide
  • 00:23:32
    you can see that as the human editor I
  • 00:23:36
    have now come in after quillbot and I've
  • 00:23:38
    added some stance features so instead of
  • 00:23:42
    repeating students and faculty or
  • 00:23:45
    Educators and students sorry I changed
  • 00:23:48
    the word to stakeholders stakeholders
  • 00:23:51
    has a little bit of a stronger meaning
  • 00:23:53
    behind it it shows that people value
  • 00:23:56
    this and that they they have a stake and
  • 00:23:58
    what's going on instead of saying they
  • 00:24:01
    should I've changed to it is critical to
  • 00:24:04
    carefully consider so you can say that
  • 00:24:06
    I'm like enhancing the language here
  • 00:24:09
    with stance features and again if you're
  • 00:24:11
    not sure what stance features are you
  • 00:24:13
    can go to that webinar and you can watch
  • 00:24:15
    me dive deep into what are these
  • 00:24:17
    features and how do I get them in my
  • 00:24:18
    writing you could also ask an AI to help
  • 00:24:22
    you understand stance and to generate
  • 00:24:25
    some ideas for how you could inject a
  • 00:24:28
    stronger or a weaker stance in your
  • 00:24:30
    writing I'm not saying it needs to be
  • 00:24:32
    strong necessarily I'm just saying you
  • 00:24:34
    need to have a control over that doubt
  • 00:24:36
    and certainty so I won't say much more
  • 00:24:39
    about that but you can see there that I
  • 00:24:41
    have improved it to some extent another
  • 00:24:44
    kind of depth that you can add inject
  • 00:24:47
    back into writing that maybe has been
  • 00:24:50
    output by Ai and this goes for I'm
  • 00:24:53
    speaking to two different audiences here
  • 00:24:55
    if you're using AI to write this is a
  • 00:24:57
    consider for you but also if you're
  • 00:25:00
    evaluating writing from as a peer
  • 00:25:03
    reviewer or as a professor of graduate
  • 00:25:06
    students or any students really and
  • 00:25:08
    you're trying to decide is this AI
  • 00:25:11
    generated or not or or what do I need to
  • 00:25:13
    say about this these are some features
  • 00:25:14
    that you can and start to figure out
  • 00:25:16
    because as I said I think we're going to
  • 00:25:17
    start seeing a lot more generated text
  • 00:25:20
    self-identification is another depth
  • 00:25:22
    feature it has to do with how the author
  • 00:25:24
    overtly inserts their identity amongst
  • 00:25:27
    the voices this is something novice
  • 00:25:29
    research writers really struggle with
  • 00:25:31
    and it has to do with their power
  • 00:25:33
    dynamic between asserting their voice in
  • 00:25:36
    the disciplinary conversation and so
  • 00:25:39
    it's really difficult to do it's hard to
  • 00:25:42
    disagree with somebody who's a seminal
  • 00:25:45
    author in the field even if you're an
  • 00:25:47
    established researcher it's difficult to
  • 00:25:48
    do that this is reflected through
  • 00:25:50
    personal perspectives which again AI
  • 00:25:53
    doesn't have and so kind you're on on
  • 00:25:56
    the up and up there because you're if
  • 00:25:58
    you're competing with an AI it doesn't
  • 00:25:59
    have any personal perspective it
  • 00:26:02
    provides connections to the writer's
  • 00:26:03
    experience or opinion and it's marked
  • 00:26:05
    with reporting verbs so if you go to the
  • 00:26:08
    next slide here's a list of ways that we
  • 00:26:11
    do this we're starting to see more
  • 00:26:13
    first-person pronouns like we and I in
  • 00:26:16
    in research writing we used to not see
  • 00:26:18
    that at all we're starting to see more
  • 00:26:20
    active voice in research writing it used
  • 00:26:22
    to air on the side more of passive voice
  • 00:26:25
    especially in the in the biomedical
  • 00:26:26
    Sciences clarify they say and I say is a
  • 00:26:30
    big one for novices and if you're not
  • 00:26:32
    familiar there's a great book called
  • 00:26:34
    they say I say and if you're training
  • 00:26:37
    beginning academic writers it's a great
  • 00:26:39
    book to have and then as I said before
  • 00:26:42
    reporting verbs which the next slide has
  • 00:26:44
    a list of for you this is one of the
  • 00:26:46
    things I tell my students and I noticed
  • 00:26:47
    some of you said no repetition and this
  • 00:26:50
    is what I tell my students I say I don't
  • 00:26:52
    want to see a reporting verb twice at
  • 00:26:55
    all I don't care if your paper is 30
  • 00:26:57
    pages there are so many reporting verbs
  • 00:27:00
    you never need to repeat one and so you
  • 00:27:03
    need to decide do you want to be
  • 00:27:04
    tenative do you want to be assertive or
  • 00:27:07
    are you taking some kind of neutral
  • 00:27:09
    literature so that's just a quick
  • 00:27:11
    resource there for you and I also want
  • 00:27:13
    to chime in Kimberly that even if you're
  • 00:27:16
    not using any sort of AI generated text
  • 00:27:19
    which we're not necessarily promoting we
  • 00:27:21
    just have different audiences here you
  • 00:27:23
    can use AI to help you evaluate your
  • 00:27:26
    writing for this you can say hey I wrote
  • 00:27:28
    this is it coming across as neutral or
  • 00:27:30
    assertive or help me identify all of my
  • 00:27:32
    reporting verbs and then categorize them
  • 00:27:34
    and maybe visually you see that most of
  • 00:27:36
    your reporting verbs are in the neutral
  • 00:27:38
    category and then you can be really
  • 00:27:40
    intentional about how you go in and
  • 00:27:42
    revise it and check your claims to see
  • 00:27:44
    if you want to be more assertive or
  • 00:27:45
    maybe a little more tentative so it's
  • 00:27:47
    not just a matter of writing in this way
  • 00:27:50
    or telling the AI to write in this way
  • 00:27:52
    you can use the AI in all of these
  • 00:27:53
    situations to act as like a peer
  • 00:27:55
    reviewer to evaluate your writing for
  • 00:27:57
    these specific features of stance yes so
  • 00:28:02
    on this slide you can see I have that
  • 00:28:04
    same paragraph that I started with it's
  • 00:28:06
    the human and then we have the quillbot
  • 00:28:08
    generated text you've seen that kind of
  • 00:28:11
    hot pink or magenta colored um those
  • 00:28:15
    stance features and now I'm adding self
  • 00:28:17
    ID or authorial voice features here and
  • 00:28:20
    this is where you get into citing other
  • 00:28:23
    literature or asserting your on voice
  • 00:28:26
    and so you'll see the use of many
  • 00:28:28
    scholars have asserted and then you see
  • 00:28:30
    those first- person pronouns we and us
  • 00:28:33
    it's not much but it adds a little bit
  • 00:28:36
    to a little bit of depth that it's not a
  • 00:28:39
    difficult thing but it really SE sets
  • 00:28:41
    the writing apart from what it looked
  • 00:28:43
    like before okay engagement markers
  • 00:28:46
    engagement markers are strategies to
  • 00:28:48
    involve the reader I think a lot of
  • 00:28:50
    times we are so inside our heads as
  • 00:28:52
    writers that we forget about what the
  • 00:28:55
    reader might have as a question or where
  • 00:28:58
    the reader is coming from um certainly
  • 00:29:00
    with students this is an issue and
  • 00:29:03
    really shifting their attention to
  • 00:29:04
    audience is an important aspect of their
  • 00:29:07
    development as academic writers um but
  • 00:29:10
    engagement markers make the writing more
  • 00:29:11
    interactive by addressing the readers
  • 00:29:14
    anticipating their knowledge or their
  • 00:29:16
    reactions and noting where something is
  • 00:29:19
    maybe surprising or unfortunate or
  • 00:29:23
    disappointing maybe your hypothesis
  • 00:29:26
    didn't turn out and you're surprised at
  • 00:29:27
    that so this is the kind of thing that
  • 00:29:29
    I'm talking about this guides the
  • 00:29:31
    understanding or the response of the
  • 00:29:33
    reader and so on the next slide you can
  • 00:29:35
    see that I have added now self ID into
  • 00:29:40
    this quillbot generated text so that it
  • 00:29:43
    has stance it has engagement and it has
  • 00:29:47
    sorry I think I missed one it has we
  • 00:29:49
    didn't this should say engagement at the
  • 00:29:51
    top of the slide not self ID that should
  • 00:29:53
    say engagement and so if you'll see in
  • 00:29:56
    the middle there this blue have led to
  • 00:29:58
    practices the reader is likely familiar
  • 00:30:00
    with such as bans on AI the use of AI
  • 00:30:03
    detection humanization tools
  • 00:30:05
    demonstrating how people might misuse
  • 00:30:08
    the technology and then there's like a
  • 00:30:09
    rhetorical question dropped in there so
  • 00:30:12
    how will higher education adapt so it
  • 00:30:15
    just makes the little bit more
  • 00:30:16
    interactive and I think we shy away from
  • 00:30:18
    that as academic writers and we think it
  • 00:30:20
    needs to be very dry but I think that's
  • 00:30:23
    what makes us our human writing stand
  • 00:30:25
    out and so I encourage you to learn a
  • 00:30:28
    little bit more about engagement and
  • 00:30:31
    start to use it in your writing and so
  • 00:30:34
    what we really have here is we have some
  • 00:30:36
    promises and some marketing kind of
  • 00:30:39
    language and then we have reality
  • 00:30:41
    basically all of these companies are
  • 00:30:43
    going to promise some form of
  • 00:30:45
    productivity enhancement in the way of
  • 00:30:48
    faster better quicker all of those
  • 00:30:51
    things they will say polished they will
  • 00:30:53
    say easier and maybe those things are
  • 00:30:56
    true but at what are at what level are
  • 00:31:00
    you achieving that is it at a level of
  • 00:31:03
    depth or are you just scratching the
  • 00:31:05
    surface here's a couple of examples down
  • 00:31:07
    at the bottom elicit which is a lit
  • 00:31:09
    Search tool they advertise superhuman
  • 00:31:13
    speed and I'm not sure if you can see
  • 00:31:15
    the one at for consensus down at the
  • 00:31:17
    bottom but it says find and understand
  • 00:31:19
    the best science faster not sure we want
  • 00:31:23
    to beat up the re certainly individually
  • 00:31:27
    yes I want to speed up the parts that I
  • 00:31:30
    don't like and get to the stuff that I
  • 00:31:32
    do but scientific rigor and speed don't
  • 00:31:38
    exactly go together and Jessica and I
  • 00:31:41
    have some concerns about this shifting
  • 00:31:44
    over to the reality the loss of context
  • 00:31:47
    because context really matters the
  • 00:31:51
    perhaps decontextualization of field
  • 00:31:53
    expertise so an AI is not a disciplinary
  • 00:31:56
    expert in in any way the fact that
  • 00:31:59
    quality varies so widely if you tried
  • 00:32:02
    chat GPT when it was in version 3.5 and
  • 00:32:05
    you've tried it now there's a huge
  • 00:32:07
    difference if you haven't if you didn't
  • 00:32:09
    try it then and compare it to now you
  • 00:32:11
    should because it's gotten marketly
  • 00:32:14
    better and then of course there are some
  • 00:32:16
    ethical
  • 00:32:17
    considerations yeah let's talk about
  • 00:32:19
    ethical
  • 00:32:21
    considerations simplification bias is
  • 00:32:24
    something that Kimbrell and I have
  • 00:32:26
    recently been digging into to a bit more
  • 00:32:29
    and what that really just means is like
  • 00:32:31
    the tendency to oversimplify something
  • 00:32:33
    that's complex leading yourself or even
  • 00:32:37
    the reader to an inaccurate or
  • 00:32:39
    incomplete picture and we stumbled upon
  • 00:32:42
    this because one of the first papers we
  • 00:32:45
    published last year the first paper we
  • 00:32:47
    did publish on generative AI has been
  • 00:32:49
    cited quite a bit and now research gate
  • 00:32:51
    has a feature that uses AI where it
  • 00:32:53
    sends me Snippets of how that study has
  • 00:32:57
    beened by other Scholars it'll highlight
  • 00:33:00
    the sentence and as I see those Snippets
  • 00:33:02
    come through maybe half of them have
  • 00:33:04
    been incorrect theyve they've made an
  • 00:33:07
    incorrect Claim about our research and
  • 00:33:09
    so it made Kimberly and I curious we
  • 00:33:11
    know that this has always been an issue
  • 00:33:13
    in research but we're concerned that AI
  • 00:33:17
    with its it's often summarizes really
  • 00:33:20
    complex papers multiple at a time and
  • 00:33:22
    produces this neat little summary for
  • 00:33:24
    you most of the AI research assistants
  • 00:33:27
    do this if you upload an article into
  • 00:33:29
    chat GPT or Claud and say summarize this
  • 00:33:32
    you are very likely going to see some
  • 00:33:35
    degree of simplification bias in what is
  • 00:33:38
    returned to you and this is an ethical
  • 00:33:40
    consideration especially when we think
  • 00:33:42
    about the context of healthc care so
  • 00:33:45
    ultimately this is like a context crisis
  • 00:33:48
    through the simplification this quick
  • 00:33:50
    summary of information we are losing the
  • 00:33:53
    context and the complexity because the
  • 00:33:56
    focus becomes on optimizing ing reducing
  • 00:33:59
    friction we miss that crucial Nuance
  • 00:34:02
    we're stripping away context and yet AI
  • 00:34:06
    generates really plausible sounding text
  • 00:34:08
    because it's very confident I think
  • 00:34:10
    that's partly why some people will get
  • 00:34:12
    this AI generated text and they don't
  • 00:34:14
    question it I have even Fallen prey to
  • 00:34:17
    that I recently started using a tool
  • 00:34:19
    that's being beta tested by Stanford
  • 00:34:22
    University it's called storm and they're
  • 00:34:25
    doing research on it so it immediately
  • 00:34:27
    ask ask for you to consent for it to
  • 00:34:29
    track your your chats and because of the
  • 00:34:32
    name Stanford it definitely lends itself
  • 00:34:34
    to some credibility that's a really
  • 00:34:37
    prestigious institution here in the US
  • 00:34:39
    and you just give it a question and
  • 00:34:42
    it'll generate an entire research
  • 00:34:44
    article free with hyperlinks to sources
  • 00:34:47
    I immediately noticed that over half of
  • 00:34:49
    them were linked to sources that were
  • 00:34:51
    either completely unrelated to the claim
  • 00:34:53
    that was being made or somewhat
  • 00:34:56
    tangentially related that the claim was
  • 00:34:59
    overstepping in some way and so this is
  • 00:35:02
    the simplification bias in real form and
  • 00:35:05
    it's like an AI context crisis that
  • 00:35:08
    we're very concerned about and so
  • 00:35:10
    Kimberly added this ENT I'm going to let
  • 00:35:12
    her explain this because she's the one
  • 00:35:13
    who's a linguist and she's familiar with
  • 00:35:15
    the linguistic framework Theory
  • 00:35:17
    relevance Theory yeah so there there's
  • 00:35:20
    this relevance theory in from
  • 00:35:22
    Linguistics and it's really more like a
  • 00:35:25
    philosophy that communication relies on
  • 00:35:28
    a shared context and for example if I
  • 00:35:31
    were to call my husband right now and
  • 00:35:33
    say it's raining he would know that mean
  • 00:35:36
    probably means something like I don't
  • 00:35:37
    have a raincoat or I don't have an
  • 00:35:39
    umbrella and I've got to walk to my car
  • 00:35:41
    because I'm at work or something like
  • 00:35:42
    that but you would not know what in the
  • 00:35:45
    world I was meaning with it's reigning
  • 00:35:48
    because you don't have any shared
  • 00:35:49
    context with me not for that situation
  • 00:35:53
    and it's the same in scientific writing
  • 00:35:56
    people in the same discipline have a
  • 00:35:58
    shared context of methodology of
  • 00:36:01
    benchmarks for evaluating certain items
  • 00:36:05
    variables that you're researching and so
  • 00:36:08
    oftentimes as experts we will omit that
  • 00:36:10
    obvious information when we're talking
  • 00:36:13
    to somebody in our discipline if you are
  • 00:36:15
    reading your journal and it's in your
  • 00:36:18
    discipline you'll notice probably that
  • 00:36:20
    the methodology there may be some things
  • 00:36:23
    that are left out however if you are
  • 00:36:26
    submitting something to an interdiscip
  • 00:36:27
    AR Journal you have to be more clear
  • 00:36:29
    about that because not everybody is
  • 00:36:31
    sharing that same context as you with an
  • 00:36:34
    AI there's no shared context whatsoever
  • 00:36:38
    and so a lot of things are getting lost
  • 00:36:41
    and we're calling that simplification
  • 00:36:42
    bias but why do our brains accept this
  • 00:36:45
    why are these AI companies taking off
  • 00:36:48
    selling these ideals around speed and
  • 00:36:51
    efficiency and optimization it's because
  • 00:36:53
    from a psychologist lens like our brains
  • 00:36:56
    want to accept we naturally seek
  • 00:36:59
    efficiency we fill in the blanks we make
  • 00:37:01
    assumptions constantly and we prefer
  • 00:37:04
    clear simple narratives that's just
  • 00:37:07
    human nature and AI tools definitely
  • 00:37:10
    capitalize on these Tendencies so it's
  • 00:37:13
    something to be aware of it does not
  • 00:37:15
    mean that you don't use or you reject
  • 00:37:17
    the technology it just means that when
  • 00:37:19
    you go to use it because we use a ton of
  • 00:37:22
    tools on a daily basis on any given day
  • 00:37:24
    I'm shuffling between three or four
  • 00:37:26
    different aiol tools but I have all of
  • 00:37:29
    this in mind I don't take everything at
  • 00:37:31
    face value I'm highly critical of what
  • 00:37:34
    it says I don't just take everything as
  • 00:37:37
    truth even though sometimes it's very
  • 00:37:39
    tempting to because it would certainly
  • 00:37:40
    save me time so when I use these tools
  • 00:37:43
    when Kimberly and I use them when we
  • 00:37:46
    talk about using them is not for Speed
  • 00:37:48
    or efficiency but for Quality we really
  • 00:37:50
    do believe that when you learn to
  • 00:37:51
    leverage this technology ethically and
  • 00:37:54
    responsibly that you can produce higher
  • 00:37:57
    quality ideas and higher quality
  • 00:37:59
    research and so what's happening right
  • 00:38:01
    now that we're seeing is like this
  • 00:38:02
    perfect storm where we have all of these
  • 00:38:04
    AI tools that are optimized for core
  • 00:38:08
    meanings and then human cognitive bias
  • 00:38:10
    Like We crave simplification and there's
  • 00:38:13
    always this pressure for research
  • 00:38:15
    efficiency that's happening even though
  • 00:38:17
    Kimberly and I are not on a tenure track
  • 00:38:19
    we feel it because we're in a field
  • 00:38:20
    that's advancing so rapidly we don't
  • 00:38:22
    want to get left behind and so the
  • 00:38:25
    result is this systematic loss of really
  • 00:38:27
    important context and the only way to
  • 00:38:31
    combat this perfect storm to not get
  • 00:38:33
    lost in it is to have all of these
  • 00:38:36
    limitations in mind and to be really
  • 00:38:38
    thoughtful and intentional about what
  • 00:38:40
    tools you're using and how and that
  • 00:38:42
    requires a bit more time actually it
  • 00:38:45
    might add on time to your normal process
  • 00:38:47
    in order to start learning to engage
  • 00:38:50
    with these tools I would argue that it's
  • 00:38:52
    worth that time because we are
  • 00:38:54
    increasingly in an AI integrated world
  • 00:38:57
    and you need to use it in order to
  • 00:38:59
    understand its capabilities and
  • 00:39:00
    limitations but please don't forget this
  • 00:39:03
    importance of context so here's just an
  • 00:39:06
    example of what this might look like
  • 00:39:07
    across disciplines so an AI or you might
  • 00:39:10
    read a sentence that says we used a
  • 00:39:12
    mixed methods approach if you're coming
  • 00:39:14
    from the discipline of
  • 00:39:16
    psychology or if you're reading a
  • 00:39:18
    journal about measures of psychological
  • 00:39:20
    validity different scales you might
  • 00:39:22
    inherently know that means qualitative
  • 00:39:24
    surveys Quant surveys and qualitative
  • 00:39:27
    energy
  • 00:39:27
    but someone in environmental science who
  • 00:39:29
    reads that sentence might immediately
  • 00:39:30
    think of field sampling and laboratory
  • 00:39:33
    analysis because that context is envir
  • 00:39:35
    environmental variable interactions and
  • 00:39:37
    in public health your mind might go to
  • 00:39:40
    epidemiological data and Community
  • 00:39:42
    feedback because you're thinking about
  • 00:39:43
    population health determinance and this
  • 00:39:46
    is where AI falls short it cannot put
  • 00:39:49
    itself in your shoes as a disciplinary
  • 00:39:53
    expert and so that's where we have to
  • 00:39:55
    give it that context in order to get the
  • 00:39:57
    most out of our experience with AI we
  • 00:39:59
    have to add in all of that context back
  • 00:40:02
    to it and so that's why often times
  • 00:40:04
    Kimberly and I when we write prompts
  • 00:40:06
    they might be several pages long and
  • 00:40:09
    that takes a lot of time and through
  • 00:40:11
    writing those prompts we actually find
  • 00:40:13
    that we get a lot more clarity about
  • 00:40:15
    what it is we're trying to say and what
  • 00:40:17
    we're evaluating and what we're looking
  • 00:40:18
    for but when you're taking in the output
  • 00:40:22
    from an AI tool you have to remember
  • 00:40:26
    that there's an an audience who's the
  • 00:40:28
    audience is it you is it someone else
  • 00:40:30
    and what is the context that might be
  • 00:40:32
    missing from that sentence or that
  • 00:40:34
    paragraph So when we think of the
  • 00:40:36
    application to research practice think
  • 00:40:39
    about these things like what
  • 00:40:40
    disciplinary context is assumed and this
  • 00:40:43
    is one of the most common ways that I
  • 00:40:44
    use AI is I ask for it to challenge what
  • 00:40:48
    assumptions I'm making in my writing
  • 00:40:51
    then you think about what methodological
  • 00:40:53
    Details Matter What implementation
  • 00:40:56
    factors are really based on your sample
  • 00:40:58
    your population um thinking about
  • 00:41:00
    ethical considerations in research and
  • 00:41:03
    then what are the field specific
  • 00:41:04
    standards that apply and all of these
  • 00:41:07
    together if you're using an AI and you
  • 00:41:09
    have all of this in mind then we do
  • 00:41:11
    believe that you can navigate this AI
  • 00:41:13
    landscape and use it ethically and
  • 00:41:15
    responsibly for your research and your
  • 00:41:17
    writing and then there's citation
  • 00:41:19
    acknowledgement this is an ethical
  • 00:41:21
    consideration as Kimberly mentioned
  • 00:41:23
    earlier we do not support the use of AI
  • 00:41:25
    detectors because they are are very
  • 00:41:28
    unreliable we talk a lot about
  • 00:41:31
    acknowledgment and citation and the
  • 00:41:33
    difference here is when you site an AI
  • 00:41:37
    tool you would site it in the same way
  • 00:41:39
    that you would any other software so
  • 00:41:41
    typically that comes up maybe in your
  • 00:41:43
    methods we cited the use of chat GPT in
  • 00:41:46
    our methods we had to say what version
  • 00:41:49
    it was because we used it to analyze
  • 00:41:51
    text in a corpus just like you would if
  • 00:41:53
    you were using maybe invivo or SPSS so
  • 00:41:57
    that how you cite it maybe as a method
  • 00:42:00
    acknowledgement is where you're
  • 00:42:02
    acknowledging its role in the process
  • 00:42:03
    and this is becoming increasingly common
  • 00:42:06
    the several the last several articles we
  • 00:42:08
    have published we have added an
  • 00:42:10
    acknowledgement of how AI was used in
  • 00:42:13
    our word and so I encourage everyone to
  • 00:42:15
    just get in the habit of reflecting on
  • 00:42:18
    their use of AI and start documenting it
  • 00:42:21
    and start to learn you'll learn more
  • 00:42:23
    about yourself and your process and what
  • 00:42:24
    works for you and what doesn't and then
  • 00:42:27
    the added benefit is you'll already have
  • 00:42:28
    that information for an acknowledgement
  • 00:42:31
    if it is indeed allowed for the venue
  • 00:42:33
    you're submitting that text to and then
  • 00:42:36
    finally data privacy this is really
  • 00:42:38
    important in healthcare I work with
  • 00:42:40
    healthc care providers these ideas
  • 00:42:43
    around Hippa and Phi still very much
  • 00:42:46
    apply I do not recommend at all entering
  • 00:42:50
    any identifiable protected health
  • 00:42:53
    information into an AI that would likely
  • 00:42:56
    very much be a breach of confidentiality
  • 00:42:59
    and data data privacy even data that is
  • 00:43:02
    anonymized I do not think it's ethical
  • 00:43:05
    to do that then you think about research
  • 00:43:07
    data protection if you're working on a
  • 00:43:10
    study and you have a novel idea you want
  • 00:43:13
    to make sure that new research data that
  • 00:43:15
    you have whether or not is this is Phi
  • 00:43:18
    aside is protected and is not going to
  • 00:43:20
    be identified by the organization that
  • 00:43:24
    the company that's created the tool that
  • 00:43:26
    you're using and so really getting in
  • 00:43:28
    the habit of looking at the data privacy
  • 00:43:31
    and security policies for the companies
  • 00:43:33
    that with the tools that you're using
  • 00:43:36
    and then that goes into just specific
  • 00:43:37
    tool considerations like many large
  • 00:43:39
    language models allow you to turn off
  • 00:43:41
    your chat history and training that
  • 00:43:44
    means that the next time you log in that
  • 00:43:46
    chat will not be there it disappears as
  • 00:43:48
    soon as you log out um it's important
  • 00:43:50
    for you to understand and start to do
  • 00:43:52
    some homework on the different AI tools
  • 00:43:54
    you're using to understand what's
  • 00:43:56
    happening with your text and your data
  • 00:43:59
    this is more critical than ever and it
  • 00:44:02
    needs to be top of mind when you're
  • 00:44:04
    using any of these tools so we have some
  • 00:44:06
    final moving forward recommendations and
  • 00:44:08
    then we're going to have a conversation
  • 00:44:10
    with you so this is all about
  • 00:44:12
    augmentation not automation so you're
  • 00:44:14
    thinking of these tools as supplements
  • 00:44:16
    not replacements for your expertise
  • 00:44:19
    you're continually maintaining deep
  • 00:44:21
    engagement with the primary literature
  • 00:44:23
    so don't let the AI do the sighting for
  • 00:44:26
    you or the synthesis for you actively
  • 00:44:29
    seek out missing context assume that
  • 00:44:32
    there's missing context and then look
  • 00:44:34
    for it this is part of critical AI
  • 00:44:36
    literacy really critically examine
  • 00:44:38
    summaries for oversimplification because
  • 00:44:40
    it is out there all of these AI research
  • 00:44:43
    assistants are trying to give you a
  • 00:44:45
    quick simple answer to a very complex
  • 00:44:48
    problem and they're oversimplifying the
  • 00:44:50
    sources they're pulling information from
  • 00:44:52
    and then document that context really
  • 00:44:55
    explicitly in your writing i' say say
  • 00:44:57
    even more so than you're perhaps used to
  • 00:45:00
    that context is really key in this AI
  • 00:45:03
    integrated world that we're finding
  • 00:45:05
    ourselves in so that is all I'm going to
  • 00:45:08
    stop sharing I'd love to just be able to
  • 00:45:10
    look at some faces and see what's in the
  • 00:45:12
    chat and just have a discussion with the
  • 00:45:15
    few minutes that we have left thank you
  • 00:45:18
    so much Jessica um I kimly that was
  • 00:45:22
    super interesting so there's a couple of
  • 00:45:25
    questions that I popped in the chat but
  • 00:45:27
    everyone else feel free to jump on and
  • 00:45:29
    with or to just unmute and turnera on to
  • 00:45:32
    we yeah call free to unmute we love
  • 00:45:35
    talking and happy to read so one of the
  • 00:45:39
    questions I had was around using good
  • 00:45:41
    fronts because it I think when I first
  • 00:45:43
    started using like co-pilot and chat
  • 00:45:46
    GPT I thought it was like Google so I'd
  • 00:45:49
    just type in a question and actually the
  • 00:45:53
    more I use it the more I realize you
  • 00:45:55
    have to get really specific and like you
  • 00:45:57
    said the prompts you can be asking it to
  • 00:45:59
    write one paragraph and you end up
  • 00:46:01
    writing two paragraphs to get it to do
  • 00:46:03
    the right thing but yeah so I was asking
  • 00:46:06
    about resources and so Kim's very
  • 00:46:08
    helpfully put a YouTube video link there
  • 00:46:10
    which is great yeah Alison to your point
  • 00:46:13
    though we've been saying the word
  • 00:46:15
    context but now I'm going to use it in a
  • 00:46:17
    different way the reason many
  • 00:46:20
    universities are struggling with this is
  • 00:46:22
    because they want to create like an
  • 00:46:23
    umbrella policy around AI use and it
  • 00:46:26
    doesn't make sense it is highly
  • 00:46:28
    contextual how it should be used I think
  • 00:46:31
    about like my first year first semester
  • 00:46:33
    doctoral students I don't want them
  • 00:46:36
    jumping in and using generative AI right
  • 00:46:38
    away so I scaffold it whereas my final
  • 00:46:41
    year students who done their original
  • 00:46:43
    research and they're getting ready to
  • 00:46:46
    defend I give them a little bit more
  • 00:46:48
    leeway because they have now become more
  • 00:46:51
    of an expert in the discipline and so
  • 00:46:54
    it's all about the level of the learner
  • 00:46:56
    if from a learning context and what are
  • 00:46:59
    the learning outcomes it is making us
  • 00:47:02
    rethink assessments but then if you're
  • 00:47:04
    an expert I find that when you're an
  • 00:47:07
    expert in your discipline and you
  • 00:47:08
    already know the literature it makes it
  • 00:47:10
    so much easier for you to spot those
  • 00:47:13
    inaccuracies way faster than someone who
  • 00:47:15
    doesn't know the literature and I think
  • 00:47:17
    it is perfectly fine for you to maybe
  • 00:47:19
    start with some machine in the loop and
  • 00:47:21
    have it generate something I struggle
  • 00:47:22
    with writer blog but then I go in and
  • 00:47:25
    heavily edit it and might not look
  • 00:47:26
    anything like what the original output
  • 00:47:28
    was and so it is really dependent upon
  • 00:47:31
    the situation and I really liked the
  • 00:47:34
    idea that you don't have to use it to
  • 00:47:36
    generate writing you can use it to
  • 00:47:38
    evaluate your own writing and that idea
  • 00:47:41
    of being really specific and say find
  • 00:47:43
    all the what they were called reporting
  • 00:47:46
    verbs and classify them and tell me what
  • 00:47:49
    tone I've taken here is really cool
  • 00:47:51
    because that's the sort of thing when if
  • 00:47:54
    you haven't studied Linguistics and
  • 00:47:56
    writing often you don't even I've never
  • 00:47:58
    even heard of a reporting verb so to be
  • 00:48:01
    able to go through and a bit more
  • 00:48:03
    deliberate about some of those things
  • 00:48:05
    almost with kind of an AI writing coach
  • 00:48:07
    it's really I think that's really
  • 00:48:10
    exciting way ofing yeah for
  • 00:48:14
    Moxy yeah feedb that's basically what
  • 00:48:18
    Moxy does is you come to the table with
  • 00:48:20
    some academic writing and it's a
  • 00:48:22
    feedback coach you can even use it if
  • 00:48:25
    you notice a piece of writing that
  • 00:48:28
    really stands out to you and you're like
  • 00:48:29
    ah this reads so well put it into an AI
  • 00:48:34
    and ask it to break it down for you in
  • 00:48:36
    terms of academic writing Concepts like
  • 00:48:38
    what makes this a good piece of writing
  • 00:48:40
    help me understand and It'll point
  • 00:48:41
    things out to you like notice these
  • 00:48:44
    transitions these reporting wordss and
  • 00:48:45
    then you start to expose yourself to New
  • 00:48:47
    Concepts and you can start to emulate
  • 00:48:49
    that in your own writing yeah that's a
  • 00:48:51
    great idea anyone in the room have any
  • 00:48:55
    questions oh so I was wondering oh
  • 00:49:00
    online asked about do you recognize
  • 00:49:02
    people using AI in their writing even if
  • 00:49:05
    they didn't acknowledge it or side it so
  • 00:49:07
    can you tell yeah that we were working
  • 00:49:10
    with a student who had submitted as the
  • 00:49:12
    doctoral student who had submitted some
  • 00:49:15
    work and she had failed the assignment
  • 00:49:17
    and didn't understand why she had failed
  • 00:49:19
    because it has this surface level kind
  • 00:49:22
    of synthetic accuracy to it and no depth
  • 00:49:25
    and so we that's actually how we got the
  • 00:49:28
    idea for um that webinar about um how to
  • 00:49:31
    assess writing without AI detectors
  • 00:49:33
    because on the surface you may or may
  • 00:49:36
    not be able to tell students certainly
  • 00:49:39
    can't tell if they have not read a lot
  • 00:49:40
    of research writing published research
  • 00:49:42
    writing I personally can tell typically
  • 00:49:47
    but it's going to get better and better
  • 00:49:49
    just because you want to weigh in on
  • 00:49:50
    that and that would be presumably only
  • 00:49:53
    if they've just used the content and
  • 00:49:57
    copied and pasted whereas if you've gone
  • 00:49:59
    through the process like you were
  • 00:50:01
    demonstrating I put in my prompt I've
  • 00:50:03
    got some outputs and now I'm refining it
  • 00:50:06
    based on my expertise and my
  • 00:50:08
    writing then it becomes possible to tell
  • 00:50:12
    that it's just given you a sort of
  • 00:50:13
    Fairly preppy rough this draft Yeah if
  • 00:50:15
    you think about that paragraph that was
  • 00:50:17
    transformed by the time Kimberly ended
  • 00:50:18
    with maybe there are a few phrases that
  • 00:50:20
    were from the original quillbot
  • 00:50:22
    generated text like that would be
  • 00:50:24
    difficult to tell and this specific
  • 00:50:27
    scenario that Kimberly's talking about
  • 00:50:28
    it was a student the assignment required
  • 00:50:30
    some reflection and connection to their
  • 00:50:32
    practice setting which was completely
  • 00:50:34
    missing and so that made it a lot more
  • 00:50:36
    obvious so it it does depend on what's
  • 00:50:39
    being written but I I think importantly
  • 00:50:42
    like I have tried to create a culture of
  • 00:50:44
    transparency in my classroom because I
  • 00:50:46
    need to learn and understand how my
  • 00:50:48
    students are using it and I just have an
  • 00:50:51
    open conversation if I suspect it and I
  • 00:50:53
    just say I just need you to walk me
  • 00:50:55
    through your logic you're thinking and
  • 00:50:57
    how you came to these ideas and yeah
  • 00:51:00
    even my Admissions Office was talking
  • 00:51:01
    about how a lot of the admissions essays
  • 00:51:03
    this year were seemed suddenly all very
  • 00:51:07
    well written compared to years very well
  • 00:51:10
    written no grammatical errors or pump
  • 00:51:12
    and erors but just really
  • 00:51:15
    shallow yeah and Angy did you have a
  • 00:51:18
    question I know sorry just trying to
  • 00:51:20
    find the mute button no thanks I've just
  • 00:51:22
    been listening and enjoying it okay
  • 00:51:25
    sorry I know
  • 00:51:27
    your camera popped up for some reason oh
  • 00:51:29
    that's weird unintentional sorry but no
  • 00:51:32
    that's fine when we're at 10:00 so I
  • 00:51:35
    might wrap up there but I was just going
  • 00:51:37
    to ask one final question is there
  • 00:51:39
    something you would recommend as a good
  • 00:51:41
    sort of First Step the some who perhaps
  • 00:51:43
    hasn't beening AI or has been using it
  • 00:51:46
    like Google who's curious about using it
  • 00:51:49
    for their writing what would be have the
  • 00:51:51
    first Tas you might about suggest they
  • 00:51:54
    do I think it depends on what you're
  • 00:51:56
    struggling with I'm really big on using
  • 00:52:00
    it intentionally for a paying point so
  • 00:52:02
    if you struggle with writer block jump
  • 00:52:04
    in and and brainstorm outlines for your
  • 00:52:07
    writing if you already have a draft and
  • 00:52:09
    you struggle with
  • 00:52:10
    revision then feed it the draft and ask
  • 00:52:14
    it to give you feedback on something
  • 00:52:16
    organization structure whatever it is
  • 00:52:19
    maybe some of your Tendencies are in
  • 00:52:20
    your writing your argument your thesis
  • 00:52:22
    statement I think asking for feedback is
  • 00:52:24
    a good one um because you're not asking
  • 00:52:26
    it to rewrite for you you can even put
  • 00:52:28
    in your prompt do not start correcting
  • 00:52:30
    these issues just give me feedback but I
  • 00:52:33
    think brainstorming and ideating is a
  • 00:52:35
    really safe beginning place I never get
  • 00:52:38
    writers block anymore I used to really
  • 00:52:40
    procrastinate writing tasks because I
  • 00:52:42
    just could not sit down and look at that
  • 00:52:44
    blank page and that just doesn't happen
  • 00:52:46
    I feel like I have camera leave aside me
  • 00:52:48
    all the time I'm just like brainstorming
  • 00:52:50
    ideating constantly I think that's a
  • 00:52:52
    really loow hanging so I know we've
  • 00:52:54
    recorded the sessions for those who were
  • 00:52:56
    able to join in person uh and who are
  • 00:52:59
    listening to the recording please feel
  • 00:53:01
    free to send through questions to me and
  • 00:53:03
    we can um have a little bit of an FAQ
  • 00:53:05
    document perhaps that goes along and
  • 00:53:08
    I'll circulate the link around school
  • 00:53:11
    join today yeah we'll make a copy of the
  • 00:53:14
    resources and the presentation PDF yeah
  • 00:53:17
    and thank you for having us this was an
  • 00:53:19
    honor to talk to all of you we're really
  • 00:53:21
    passionate about this work and we put
  • 00:53:24
    out a lot of content on YouTube that's a
  • 00:53:26
    very educational so feel free to follow
  • 00:53:28
    us and check out our other webinars if
  • 00:53:30
    you want to take a deeper dive into any
  • 00:53:32
    of these topics
الوسوم
  • AI
  • academic writing
  • productivity
  • ethics
  • Moxy
  • Lit Maps
  • simplification bias
  • data privacy
  • authorial identity
  • engagement markers