Difference between revisions of "User:Ankhan4"
(→Week 7: (07/10/23 - 07/14/23)) |
(→Week 7: (07/10/23 - 07/14/23)) |
||
| Line 210: | Line 210: | ||
'''''Monday 7/10''''' | '''''Monday 7/10''''' | ||
*Worked on website draft proposal | *Worked on website draft proposal | ||
| + | *Drafted questions and ideas for website design for meeting with grad student | ||
| + | *Researched openAI API policy on inappropriate topics and moderation to figure out why it sometimes engaged with topics it deemed inappropriate prior | ||
| + | |||
| + | a) Makes it clear there are certain ways to “trick” the AI to engage with the topic, talks about subject in a very vague but unambiguous manner | ||
| + | |||
| + | 1. https://platform.openai.com/docs/guides/safety-best-practices | ||
| + | ''Explains their best practices to determine how inputs can be adjusted'' | ||
| + | |||
| + | 2. https://platform.openai.com/docs/api-reference/moderations/create | ||
| + | ''Able to prevent certain topics from being engaged with'' | ||
| + | |||
| + | 3. https://medium.com/geekculture/a-deep-dive-into-ai-safety-and-ethics-utilizing-openais-moderation-apis-4d6edb91afb4#:~:text=OpenAI's%20Moderation%20APIs%20categorize%20content,filtering%20or%20blocking%20harmful%20content. | ||
| + | ''Resource that discusses application of open ai moderation api and how it can help prevent ethical violations'' | ||
Revision as of 17:20, 10 July 2023
Contents
About Me
Hi! My name is Amal Khan and I am a senior at UW-Madison studying Philosophy and Data Science.
Work Log
Week 1: (05/30/23 - 06/02/23)
Tuesday 5/30
- Attended the REU Introduction
Wednesday 5/31
- Met with mentors to understand goal of project
- Reviewed previously published material from lab
1. Progga, Kumar, Rubya “Understanding the Online Social Support Dynamics for Postpartum Depression”, 2023 https://dl.acm.org/doi/abs/10.1145/3565967.3570977
In depth discussion about findings regarding topics talked about in PPD forums and the types of support that mothers receive from forums while discussing limitations of machine learning and NLP application.
2. Rubya, Progga “‘just like therapy!’: Investigating the Potential of Storytelling in Online Postpartum Depression Communities” , 2023 https://dl-acm-org.ezproxy.library.wisc.edu/doi/pdf/10.1145/3565967.3570977
Explains the main themes women explore through posting on PPD online health forums and how they feel supported through the online communities along with how a system that detects crisis situations in a post can provide immediate feedback to provide mothers with crisis resources and support.
Thursday 6/1
- Met with mentor to discuss milestones and discuss article topics
- Familiarized with dataset from PPD online forum
- Conducted further background research
1. Tawfiq Ammari, Sarita Schoenebeck, and Daniel Romero. 2019. Self-declared Throwaway Accounts on Reddit: How Platform Affordances and Shared Norms enable Parenting Disclosure and Support. Proc. ACM Hum.-Comput. Interact. 3, CSCW, Article 135 (November 2019), 30 pages. https://doi-org.ezproxy.library.wisc.edu/10.1145/3359237
Emphasizes benefits of anonymous accounts on online forums like Reddit and how they allow parents to talk about difficult topics without backlash or feeling exposed
2. Phil Adams, Eric PS Baumer, and Geri Gay. 2014. Staccato social support in mobile health applications. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (CHI '14). Association for Computing Machinery, New York, NY, USA, 653–662. https://doi-org.ezproxy.library.wisc.edu/10.1145/2556288.2557297
Shows how computer mediated communication can be used to boost self esteem and push participants towards healthier decisions in health through health focused app’s interactions with other users and validating follow up messages to their own posts.
Friday 6/2
- Completed online RCR training
- Uploaded project milestones
- Background research from references of previously published papers from lab
1. Baumel A, Schueller SM. Adjusting an Available Online Peer Support Platform in a Program to Supplement the Treatment of Perinatal Depression and Anxiety. JMIR Ment Health. 2016 Mar 21;3(1):e11. doi: 10.2196/mental.5335. PMID: 27001373; PMCID: PMC4820657.
Analyzed overall positive integration of e-health service similar to therapy listening service from women with perinatal depression and anxiety for women with the same health issue and how it helped them outside of the clinic but noted that though women enjoyed it they did not want to volunteer to contribute to service in future.
2. Yixin Chen and Yang Xu. 2021. Social Support is Contagious: Exploring the Effect of Social Support in Online Mental Health Communities. In Extended Abstracts of the 2021 CHI Conference on Human Factors in Computing Systems (CHI EA '21). Association for Computing Machinery, New York, NY, USA, Article 286, 1–6. https://doi-org.ezproxy.library.wisc.edu/10.1145/3411763.3451644
Noted how quality and amount of feedback on first post in online forums has causal relationship to user posting after initial post and how robots could be introduced to help mitigate overly negative or lack of responses on posts to provide more social support to users.
Week 2: (06/05/23 - 06/09/23)
Monday 6/5
- Attended in-person RCR training
- Background Research
Tuesday 6/6
- Brainstormed combination of themes for ChatGPT
- Prompted sample drafts
- Met with team to discuss findings
- Background Research regarding HCI and how to make more descriptive inputs for GPT storytelling outputs
1. Ashish Amresh, Madhumita Sinha, Rebecca Birr, and Rahul Salla. 2015. Interactive Cause and Effect Comic-book Storytelling for Improving Nutrition Outcomes in Children. In Proceedings of the 5th International Conference on Digital Health 2015 (DH '15). Association for Computing Machinery, New York, NY, USA, 9–14. https://doi-org.ezproxy.library.wisc.edu/10.1145/2750511.2750533
Created interactive game for Latino youth to bring about awareness for healthy eating habits to combat increasing rates of obesity. Noted benefits of having the main character look similar to participants and games being played by both parents and children.
2. Aleesha Hamid, Rabiah Arshad, and Suleman Shahid. 2022. What are you thinking?: Using CBT and Storytelling to Improve Mental Health Among College Students. In Proceedings of the 2022 CHI Conference on Human Factors in Computing Systems (CHI '22). Association for Computing Machinery, New York, NY, USA, Article 441, 1–16. https://doi-org.ezproxy.library.wisc.edu/10.1145/3491102.3517603
Gameified storytelling app that walked students through CBT practices noted the importance of visuals being realistic enough for users to relate and how combining games with daily self observation helped participants.
3. Hye Sun Yun, Matias Volonte, and Timothy Bickmore. 2022. Motivating health behavior change with a storytelling virtual agent. In Proceedings of the 22nd ACM International Conference on Intelligent Virtual Agents (IVA '22). Association for Computing Machinery, New York, NY, USA, Article 31, 1–3. https://doi-org.ezproxy.library.wisc.edu/10.1145/3514197.3549684
Religiously focused interactive counseling games used to improve healthy eating habits highlighted benefits of having cultural similarities between game avatar and user, reporting increased motivation to eat healthier from participants.
Wednesday 6/7
- Edited inputs for ChatGPT based on feedback from initial drafts
- Attended Technical Writing Seminar
- Weekly team meeting
Thursday 6/8
- Started draft of literature review paragraph
- Created more drafts with ChatGPT
Friday 6/9
- Finalized first batch of drafts
- Started draft of final paper
Week 3: (06/12/23 - 06/16/23)
Monday 6/12
- Selected prompts for API
- Read literature review
- Worked on paper draft
Tuesday 6/13
- Troubleshooted with API
- Worked on literature review
Wednesday 6/14
- Ran drafts for API
- Worked on literature review
- Wrote feedback for API outputs
Thursday 6/15
- Ran drafts on ChatGPT interface by feeding sample stories
- Adjusted commands
- Tried using keywords in API
Friday 6/16
- Continued working on API outputs
- Started writing methods section of paper
Week 4: (06/19/23 - 06/23/23)
Monday 6/19
- Generated more API outputs
- Wrote function that randomly selected two keywords as input for API
Tuesday 6/20
- Met with research group to discuss updates
- Generated more API outputs
- Noted limitations and patterns in output
Wednesday 6/21
- REU Student Check-in
- Met with REU student to discuss NLP processing methods
- Categorized API outputs
- Researched linguistic trends in social media
1. Pantelis Vikatos, Johnnatan Messias, Manoel Miranda, and Fabrício Benevenuto. 2017. Linguistic Diversities of Demographic Groups in Twitter. In Proceedings of the 28th ACM Conference on Hypertext and Social Media (HT '17). Association for Computing Machinery, New York, NY, USA, 275–284. https://doi-org.ezproxy.library.wisc.edu/10.1145/3078714.3078742 Analyzes linguistic trends in tweets by race and gender focusing on White, Black and Asian, male and female. Notes that certain emotions are higher in other groups that others, similarly with lexical density and interest in subject matter varies in both race and gender
2. Vedula, Nikhita, and Srinivasan Parthasarathy. “Emotional and Linguistic Cues of Depression from Social Media.” Proceedings of the 2017 International Conference on Digital Health, ACM, 2017, pp. 127–36. DOI.org (Crossref), https://doi.org/10.1145/3079452.3079465. Analyzes linguistic trends of users with depression in tweets, behaviors including using social media more in the nighttime, posting less frequently, engaging less with network, using more negative language, and utilizing self focused pronouns like “I”.
3. Swartz, Melanie, et al. “Diversity from Emojis and Keywords in Social Media.” International Conference on Social Media and Society, ACM, 2020, pp. 92–100. DOI.org (Crossref), https://doi.org/10.1145/3400806.3400818. Determine diversity composition of users in tweets surrounding the time period of 2018 midterm election based on usage of diversity emojis and keywords and reveal race, gender, sexuality and religion.
Thursday 6/22
- Started creating presentation slides
- Working on sections of paper
- Editing literature review
Friday 6/23
- Edited paper
- Presentation Work
Week 5: (06/26/23 - 06/30/23)
Monday 6/26
- Presented practice presentation with research group
- Implemented feedback into presentation
- Research articles regarding semantic analysis of LLM
1. Zhou, Jiawei, et al. “Synthetic Lies: Understanding AI-Generated Misinformation and Evaluating Algorithmic and Human Solutions.” Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems, ACM, 2023, pp. 1–20. DOI.org (Crossref), https://doi.org/10.1145/3544548.3581318. Comparing AI generated and user generated LLM finding they are significantly different and that AI had the tendency to align with four themes; enhancing details, communicating uncertainty, drawing conclusions, and human -like tone.
2. Sharma, Eva, and Munmun De Choudhury. “Mental Health Support and Its Relationship to Linguistic Accommodation in Online Communities.” Proceedings of the 2018 CHI Conference on Human Factors in Computing Systems, ACM, 2018, pp. 1–13. DOI.org (Crossref), https://doi.org/10.1145/3173574.3174215. Found that users that subscribe to linguistic trends or norms like post length, content, self disclosure, etc. of an online health community get more positive support from other users in the community.
Tuesday 6/27
- Updated related work section including new resources and increasing formality of written tone
Wednesday 6/28
- Eid-al-Adha
Thursday 6/29
- Compared findings of research articles comparing AI generated and user generated text to current project
- Determined which methods to apply and how to adjust them
Friday 6/29
- Prompted new API outputs adjusting to trends in AI generated language
Week 6: (07/03/23 - 07/03/23)
Monday 7/03
- Practiced presentation
- Worked on API function
- Brainstorming statistical analysis of findings
Tuesday 7/04
- Practiced presentation
- Prompted more API outputs
- Researched analysis methods for research questions
- Prepared questions regarding analysis methods for weekly meeting
Wednesday 7/05
- Presented mini-presentation
- Troubleshooted subject limitation issue with API
Found that API randomly chose to engage or not engage with specific keywords
- Researched SAGE analysis
1. Jacob Eisenstein, Amr Ahmed, and Eric P. Xing. 2011. Sparse additive generative models of text. In Proceedings of the 28th International Conference on International Conference on Machine Learning (ICML'11). Omnipress, Madison, WI, USA, 1041–1048. Compares distribution of certain word with background distribution of other words. Each class is a word, see how much it shows up compared to other words
Thursday 7/06
- Researched mTurk analysis method
1. https://www.mturk.com/ Service that allows you to outsource human tasks like survey participation, will be used in our project to compare AI generated posts/user generated posts
2. Hellas, Arto, et al. “Crowdsourcing in Computing Education Research: Case Amazon MTurk.” Koli Calling ’20: Proceedings of the 20th Koli Calling International Conference on Computing Education Research, ACM, 2020, pp. 1–5. DOI.org (Crossref), https://doi.org/10.1145/3428029.3428062. Discussed benefits and limitations of crowdsourcing as it provides a larger diversity of backgrounds to aid in project but is also difficult to meet necessary criteria of participants doing crowdsourced work
- Categorized API outputs
- Wrote systematic approach for API methods, listing what combinations and questions are missing
- Worked further with API to create more outputs to fill gaps to create more diverse perspectives of stories and content
Friday 7/06
- Selecting n<20 drafts to be used for website draft
- Started brainstorming website draft/search engine to understand next step
1. https://aws.amazon.com/startups/start-building/how-to-build-a-web-app/ Explained how to use AWS to create web application illustrating different steps and usage of API
2. https://budibase.com/blog/web-application-development/ Provided an outline of what questions of in planning the content and use of a web application, helpful for starting an outline
3. https://kissflow.com/application-development/how-to-create-a-web-application/ Listed steps for creating web application bringing new concepts to attention like creating a framework or minimal viable product (MVP)
Week 7: (07/10/23 - 07/14/23)
Monday 7/10
- Worked on website draft proposal
- Drafted questions and ideas for website design for meeting with grad student
- Researched openAI API policy on inappropriate topics and moderation to figure out why it sometimes engaged with topics it deemed inappropriate prior
a) Makes it clear there are certain ways to “trick” the AI to engage with the topic, talks about subject in a very vague but unambiguous manner
1. https://platform.openai.com/docs/guides/safety-best-practices Explains their best practices to determine how inputs can be adjusted
2. https://platform.openai.com/docs/api-reference/moderations/create Able to prevent certain topics from being engaged with
3. https://medium.com/geekculture/a-deep-dive-into-ai-safety-and-ethics-utilizing-openais-moderation-apis-4d6edb91afb4#:~:text=OpenAI's%20Moderation%20APIs%20categorize%20content,filtering%20or%20blocking%20harmful%20content. Resource that discusses application of open ai moderation api and how it can help prevent ethical violations
