Jump to content

2025:Program/How we can handle Wikimania scholarship applications with the invasion of AI?

Add links
From Wikimania

Session title: How we can handle Wikimania scholarship applications with the invasion of AI?

Session type: Panel
Track: Technology
Language: en

đŸŽ„ Session recording: https://w.wiki/JwS6 đŸŽ„

During this year's Wikimania scholarship review process, reviewers noticed that some applications used artificial intelligence (e.g. ChatGPT) to answer scholarship questions.

This raises many questions: Is it acceptable to use AI to answer these questions? Where do we draw the line for using it? Are AI-generated response considered to be genuine? Why do the applicants think they need to use this tool? Will it create an arms race between those who use and those who don't? Will it create an uneven playing field between applicants who can afford to pay to use paid-version of AI and those who cannot pay and have to use the free version?

In this panel discussion, we have representatives who reviewed scholarships from different regions to share their thoughts on the use of AI tools to answer Wikimania scholarship questions.

Description

The era for using artificial intelligence to answer Wikimania scholarship application questions has arrived. It raises legal, governance, ethical, moral and fairness questions over the usage. We will have a panel discussion on how AI affects the work of the scholarship committee.

Currently there is no official policy that governs the usage of AI tools to assist in answering questions on the Wikimania scholarship application. They are typically noticed when very similar responses were provided by different applicants. Those responses often lack specific examples or metrics and may misunderstand homonyms (e.g. AI misinterpreted the question with the keyword "sustainability" to mean the extraction of natural resources, energy consumption and waste disposal at the event, not the longevity of the movement). There is also no standard for handling answers that are suspected to be entirely written by AI, leading to inconsistent scoring for these applications.

With panelists from different Wikimedia regions, we hope that this discussion kickstarts a formal policy around the use of AI in scholarship response that supports non-native English speakers' participation in regional/global conferences. Wikimania participants will gain a deeper insight towards the challenges that scholarship committee members faced when reviewing and scoring Wikimania scholarships.

How does your session relate to the event theme, Wikimania@20 – Inclusivity. Impact. Sustainability?

As scholarship committee members are volunteers, there is a risk of burnout from reviewing large amount of answers that are suspected to be written by AI, threatening the sustainability of this group of volunteers. A balancing act is required to ensure genuine answers without penalizing non-native English speakers to ensure their inclusion and participation in the Wikimedia movement through travel scholarship. This issue will continue to persist until the rules governing AI usage in scholarship application is introduced.

What is the experience level needed for the audience for your session?

Everyone can participate in this session

Resources

Notes (notetaking by Netha Hussain)

(one speaker speaks French, speaking using a volunteer interpreter, one speaker joins online)

Moderator: What the session will cover:

   - Rising use of AI in Wikimania and other conferences

   - What are the perceptions of the community and the panel regarding this?

   - Where do we draw the line between using AI for scholarship applications?

   

No AI was seen used in scholarship applications in earlier conferences, but is being used more and more recently. Do review members get overwhelmed and stressed out with volumes of AI generated text?

  It is overwhelming sometimes, but volunteers somehow manage. The committee expected more AI generated text, but that was not the case. The committee was prepared to handle AI, although it was stressful and time consuming work. It took some time to find out if it was original text or AI generated text.

   

How to spot AI generated text? When do you suspect that the text is generated by AI?

  • The committee member had to go to look at the userpages and other on-wiki evidences of the users (I became a stalker! she said), so that she can corroborate if the applicant said the truth about their work on the application.

Where should we draw the line on AI?

  • We need to embrace AI, we need to set the limits, how far and safely can we use AI should be discussed. Rules and guidelines need to be established, possibly by the Foundation (WMF?). WikiIndaba scholarship checking was quite tedious, because a lot of AI text creeps in, people not responsibly using AI and copy-pasting large volumes of text.
  • There needs to be a policy for AI text, but in the end of the day, the people will need to prove what they are saying, by providing evidence for what they are claiming for in the application text.
  • Change is inevitable, we need to make changes in local and affiliate level, so that there is no unfair bias to people who use AI. It is often the unprivileged that do not have access to AI.
  • Wikimedia Foundation's CTOs approach is that we need to have three approaches : sustainability, equity and transparency.
  • We need principles to say, what is acceptable and what is not acceptable in scholarship applications, which is highly competitive process. We have 7000 applicants, only 300 scholarships.
  • Someone made 1000 edits a day, not possible for humans to do so, then it is probably AI assisted work that they are doing it.

Some countries cannot afford AI due to financial or legal regulations. So, those with higher end technology could win over those who do not have access to latest technology.

  • One way to deal with this is to ask the user to self-declare the use of AI. In a previous scholarship process, those who used AI were deprioritized over the others who did not use AI. AI is going to create an arms race, and we need to urgently develop guidelines. Communities should create AI literacy education. Wikimedia can spearhead the conversation about AI in the digital spaces. Next year, before Wikimania and WIkiIndaba, we could make them AI literate so that people use AI tools responsibly.
  • WikiIndaba this year asked a question: How did you use AI in this application? It was quite useful to know what the applications wrote in the answers.
  • The key is to give people to give evidence for what they have done for Wikimedia. You need to have access to verifiable information to show what they said is real.

In youth conference in Prague, they asked the applications to make one minute video recording to answer one of the questions, what are the solutions for AI unfairness?

  • We need to have committee members address applicants in the languages of their choice.
  • We need constant adaptation of the policies for changing landscape in AI.
  • There is gap between good intentions and actions.
  • There are people in community who are very likely eligible for a scholarship, but do not get one because of lack of confidence. Ai can probably help them articulate better. There is also a social barrier to making the first step to applying for a conference.

Comments from the audience:

  Marginalized communities have oral traditions. Call for invitations to include conversations. Balance the weight of the writing structure to voice structure, because writing is not a traditional medium for documentation in some cultures.

  differentiation between humans and AI generated text? Someone made 1000 edits a day, not possible for humans to do so, then it is probably AI assisted work that they are doing it.

  Writing things like "I love Wikipedia" and not corroborating with evidence. Text like "insert more answers here.." etc would ring a bell.

  Putting the scholarship questions into ChatGPT and know what ChatGPT says beforehand, so that one can expect what AI generated text already says.

   When you get a lot of answers of the same kind, same template, you look for originality in the text.

   In the University, most of the work is AI generated by the students.

  English is the language of the colonizer, AI is tool of the colonizer. IT is going to escalate the gap between the others and the others. Those who are not properly trained to use the tools, the gap will continue. Wikimedia is a community platform, we need to embrace technology while assisting to fill the gap.

  ESEAP conference applications: Was there ever decrease among the reviewers? What is the proportion of AI applications in absolute numbers? It is region specific, no real statistics about it. lot of uncertainity and plenty of discussions need to happen to govern the use of AI in scholarship and program applications.

Speakers

  • Andrew Leung
I am a Wikipedian since 2005. I hold administrator rights on English Wikipedia and English Wikivoyage, and bureaucrat rights on Wikispecies. I focus on Education and Wikipedia. I am the editor-in-chief for WikiJournal of Science, currently hosted in Wikiversity.
  • Herschal Jackson (WMZA)
I am the Executive Director of Wikimedia South Africa
Over the last 2 years I have worked within the chapter to implement the 2024-2027 strategy.
I believe in the goals of the chapter and the overarching Movements Goals.
From a personal perspective I am a man of great faith and believe completely in doing everything with excellence.
"What you do reflects whose you are"
  • Carol Mwaura
.Carol currently Co-Leads Wikimedia User Group Kenya where she champions for access, contribution of open free knowledge and championing the closing of gender gap on Wikipedia.Lead on projects that advocate for open knowledge and digital literacy through Wikipedia training and strategic partnership, for example Wikifor Human Rights, Art & Feminism. Participant in GLAM and WikiLibrarians communities on Wikipedia.
  • Donatien Kangah
Donatien Kangah est un WikimĂ©dien ivoirien. En 2012, il co-fonde le Groupe d'utilisateurs WikimĂ©dia CĂŽte d'Ivoire. Il collabore Ă©galement avec plusieurs communautĂ©s africaines et organise la Wikiconvention francophone 2023 Ă  Abidjan, rassemblant 130 participants. Contributeur actif sur WikipĂ©dia en français, sa philosophie repose sur l'apprentissage par le partage et le leadership comme service Ă  la communautĂ©. Il Ɠuvre pour la durabilitĂ© du mouvement en Afrique, en se concentrant sur la transition gĂ©nĂ©rationnelle et la gestion des conflits, tout en promouvant l'inclusion linguistique dans le mouvement WikimĂ©dia.
Retrieved from "https://wikimania.wikimedia.org/w/index.php?title=2025:Program/How_we_can_handle_Wikimania_scholarship_applications_with_the_invasion_of_AI%3F&oldid=547384"