00:33:44 Jennifer Delfino: Welcome to our last Town Hall of 2023! Be sure to say hello, share your linkedIn and let us know where you are from🙂 00:34:24 Chris Robson: Hi everyone from Portland! 00:34:34 Marnie Kittelson: Hi from Minneapolis! 00:34:57 Geoff Roper: Greetings from Myrtle Beach! 00:34:57 Monica Schell: Hello from Virginia! Will the recording of this event be sent to us? 00:35:02 William Mansfield: Hi from North Carolina. 00:35:05 Adam Probolsky: Hola from Newport Beach, CA 00:35:10 Damion Taylor: HI from Los Angeles! 00:35:12 Matt O'Mara: Good afternoon from sunny Detroit, MI 00:35:13 Terry Sweeney: Greetings from Rochester, NY 00:35:13 Simon Chadwick: Greetings from Raleigh, NC! 00:35:22 Dharmendra Jain: Hello Everyone from Nairobi, Kenya 00:35:23 Katie Gross: Hi from Carolina Beach, NC 00:35:30 Deborah Suckow: Hello from Cincinnati, Ohio 00:35:34 David Mills: https://www.linkedin.com/in/david-mills located in southern New Hampshire (N of Boston) seeking entry level opportunity in data analysis. Getting comfortable with Excel, Power BI, Tableau. 00:35:41 Wayne Lashua: Hello form the Northern Kentucky/Cincinnati Area! 00:35:44 Cate Crandall: Hi all! San Francisco 00:35:46 Stephen DiMarco: Hello from sunny Montclair, NJ (via Boston)! 00:35:49 Beth Coduti: Hi from Cleveland, Ohio! 00:35:49 Amanda Keller-Grill: Hi from Redondo Beach, CA! 00:35:55 Beth Horn: Hello from Dallas, TX! https://www.linkedin.com/in/bethhorn/ 00:36:00 Jen Floyd: Hi from West Virginia! 00:36:03 John Bremer: Las Vegas 00:36:07 Kelsey Miller: Hello from Cincinnati, Ohio! 00:36:13 Lilah Raynor: Hello from San Francisco! 00:36:26 Stephen DiMarco: Hi, I'm an AI assistant helping Stephen DiMarco take notes for this meeting. Follow along the transcript here: https://otter.ai/u/XuJIGCvwhN8L-2LsZtsm6hfaDHA?utm_source=va_chat_link_1 You'll also be able to see screenshots of key moments, add highlights, comments, or action items to anything being said, and get an automatic summary after the meeting. 00:36:28 Alissa Sauer: Hi from Hood River, Oregon! https://www.linkedin.com/in/alissa-ehlers-sauer-7b5bb189/ 00:36:46 Stephen DiMarco: https://www.linkedin.com/in/stephenrdimarco/ 00:36:55 Cate Crandall: Here’s my LinkedIn - would love to connect! https://www.linkedin.com/in/cate-riegner/ 00:36:58 Geoff Roper: Looking forward to this session - https://www.linkedin.com/in/geroper/ 00:37:07 Nija nair: Hello from Toronto, Canada! Experienced marketing analytics, media & consumer insights leader here ! This a very relevant topic for me :) Happy to connect https://www.linkedin.com/in/nijanair 00:37:17 Kamal Malek: Hello from Boston. https://www.linkedin.com/in/kamalmalek/ 00:37:25 Shannon Danzy (she/her): Hello everyone! This is Shannon Danzy, Director of Research from the Insights Association. https://www.linkedin.com/in/sdanzy/ 00:37:55 Dharmendra Jain: https://www.linkedin.com/in/jaindharmendra/ 00:38:29 Mary Fanning: Hi from Cincy! https://www.linkedin.com/in/marymfanning/ 00:39:27 Damion Taylor: Goo Irish!! 00:43:04 Stephen DiMarco: ☝️🤓 00:44:24 Cate Crandall: How do you “get data from models”? Can you give an example. 00:46:06 Bruce Olson: By this definition, starting point for "synthetic data" was really the application of Hierarchical Bayes to 'fill in' or 'smooth' data from respondents in DCM or Max Diff? 00:46:36 Cate Crandall: Yes, I worry about this turning into “fake news” 00:48:30 John Bremer: https://gking.harvard.edu/files/abs/not-abs.shtml 00:48:54 Leslie Christensen: Will synthetic data include outliers to mimic the real-world where there are always a few respondents, situations that don't fit a model based on "the average"? 00:53:59 John Bremer: My concern here is that when you think of this from an imputation standpoint, we quickly realized that variance wasn't really accounted for with a single imputation. We had to move to multiple imputation to get an idea of variation. Normal folks not knowing what they are doing seems to lose that nuance and in many cases it won't matter but in some really important ways it will. What are the guardrails? 00:57:19 Melanie Courtright: https://www.insightsassociation.org/Resources/Code-of-Standards 00:57:49 Melanie Courtright: Section 4: Data and Technology Researchers must apply the principles of this research Code to all elements of use in the realm of Artificial Intelligence when used for the generation of data or insights. Data that is created, curated, and utilized in Artificial Intelligence solutions must meet the criteria of data governance, duty of care, transparency, and research quality. In addition, specific to artificial intelligence, researchers must follow the principles outlined in Section 5 about the use of secondary data. 01:02:42 Akanksa Upadhyay: Is there an existing platform that generates these synthetic personas? 01:04:58 Janet Standen: Chris - If it’s based on your own unique data collected about real people from which you have created the “soccer mum persona” that stakeholders can then interact with and ask them other questions about their lives - as long as it is from the freshly collected data about this group of people - why is that considered as “synthetic” if it is rooted in primary data collection? 01:05:02 Stephen DiMarco: @katie - if you included social data (conversational data) that is up to date in the synthetic person - you could ask q's about disruptive innovations like new flavors! 01:05:55 Stephen DiMarco: (brands are using social data for product innovation now -- synthetic segments would give this practice more gravitas) 01:07:40 Manoja Manoharan: For those who incorporated synthetic data, what legal, security, and risk challenges did you encounter, and how did you navigate and overcome them? 01:11:27 Janet Standen: So everything we create in our brains / discussions as researchers based on real/primary data we have collected - is “synthetic interpretation” of real data - whether it’s done by human brains or models? 01:11:33 Freeman Lewin: How worried are the panelists about the rise of cloning? New companies are popping up from companies offering to clone voice and personalities, such that panelists in the future could, unchecked, be fake. 01:11:40 John Bremer: Think there is an issue with terminology. Can merge "synthetic data" back in to the primary data if you are using it that way, you can just create a data set by itself if that is your use, you can do x, y , and z if that is your purpose. Think having a single term we are pointing back to is going to trip us up 01:12:14 John Bremer: Why there is an issue with a precise definition because it can fit one use but not another 01:12:26 Nija nair: Thks Melanie. so does that mean that data collaborations across data/tech/agency/advertiser partners would be easier? potentially better ways to work with walled gardens, since its not exactly their data but mimics it? 01:14:04 John Bremer: Stephen, like your pov. May not fit the definition that is put out here but would be powerful. 01:15:35 Howard Fienberg: Was discussing synthetic data with Senator Schumer's staff this week... 01:15:43 John Bremer: Bigger AI question is who owns the data? Not sure there is an answer here, particularly since there is different ways of creating the data 01:16:32 Katie Gross: And that John is exactly why we have so much hesitancy from our Enterprise brand clients 01:17:44 John Bremer: Are you saying suing people solves all the issues?? 01:17:48 John Bremer: :-) 01:17:52 Melanie Courtright: NO! hahaha 01:19:33 John Bremer: @Katie yup seems to be a tough issue to crack. 01:20:21 Dharmendra Jain: What are your views about the panels which are offering synthetic data based samples? Has anyone used it? It will be great to hear about the experiences. 01:21:19 Janet Standen: What about using a platform like Yabble for this persona type purpose. 01:21:25 John Bremer: Although we are looking at this as a new issue. if you look at this as an old issue, maybe there might be some guidance. The US Census has been using an old school / simplistic version that fits the definition here for decades. Perhaps you can l;ook to that to get some idea how to handle. 01:21:34 John Bremer: That was for @Katie 01:22:41 Mary Dominiecki: Is anyone using this in healthcare market research yet? 01:23:48 Nija nair: Great to hear Zachary! I see many clients much better prepared for the future by breaking silos for transparent & holistic view of insights. 01:23:51 Jeremiah Bullock: While open-source has been big prior to the past year, it's growing in a huge way. I strongly recommend folks look into open-source tools in order to be able to experiment privately and quickly. ChromaDB (for a vector database), tools like Ollama for being able to run LLMs locally, and even LLMs such as mistral or even the newly released mixtral which uses a technique called "mixture of experts" 01:26:31 Stephen DiMarco: Action Items: [ ] Explore platforms and tools for creating synthetic personas (Chris, Damien) [ ] Consider transparency guidelines for using synthetic data (Ben, Howard) [ ] Investigate vector databases and building models with custom code (Zach, Katie) See full summary - https://otter.ai/u/XuJIGCvwhN8L-2LsZtsm6hfaDHA?utm_source=va_chat&utm_content=wrapup_v1&tab=chat&message=9d79d76f-c6b2-41ff-bdc2-6918b9d6c6d6 01:30:10 David Mills: can the chat be saved? I'd especially like to keep the link to the summary. 01:30:33 Melanie Courtright: Yes - we can deliver the chat to the group 01:30:41 David Mills: Thanks! 01:31:57 Janet Standen: The value of this session alone is a good reason to renew my IA subscription! Thanks IA and all panel members. 01:32:12 John Bremer: Would also advise to think about variation. A single draw from a probabilistic model could be made better with multiple draws 01:32:33 Stephen DiMarco: And show consumers what we are doing and how it works. If they can see it, they are less likely to feel something sneaky is going on! 01:32:34 John Bremer: @Janet, Melanie will love that. 01:32:56 Stephen DiMarco: Hello Bremer - blast from the past! 01:32:59 Cate Crandall: This has been great! Thank you! I think I’ve got a foundation of knowledge now to build on as we move forward 01:33:15 John Bremer: @Steven that is critical 01:33:23 Kamal Malek: Thank you to the panelists! 01:33:27 Bonnie Janzen: Thank you! 01:33:28 John Bremer: Good to "see you: as well. Has been forever 01:33:46 Sheri Roder: Great panel-- thanks to all! 01:33:48 Kamal Malek: And the organizers! 01:33:58 Nija nair: Thks Melanie & panelists! 01:33:58 John Bremer: Great session 01:33:59 Amanda Keller-Grill: This was great- thank you! 01:34:08 Susan Ullrich: Thank you so much! 01:34:10 Melinda Kizer: Thank you! Informative discussion. 01:34:11 Matt O'Mara: Thank you! 01:34:15 Christian Riegel: Great conversation. Thanks! 01:34:16 Dharmendra Jain: Thank you, great session! 01:34:27 Monica Schell: Thank you! Will this recording be shared?