RLP 326: Using AI with Research Planning

Nicole (1s):
This is Research Like a Pro episode 326 Using AI with Research Planning. Welcome to Research Like a Pro a Genealogy Podcast about taking your research to the next level, hosted by Nicole Dyer and Diana Elder accredited genealogy professional. Diana and Nicole are the mother-daughter team at FamilyLocket.com and the authors of Research Like a Pro A Genealogist Guide. With Robin Wirthlin they also co-authored the companion volume, Research Like a Pro with DNA. Join Diana and Nicole as they discuss how to stay organized, make progress in their research and solve difficult cases. Let’s go.

Nicole (41s):
Today’s episode is sponsored by Newspapers.com. Hi to all of our Research Like a Pro listeners, how are you all today? And hello to you Mom.

Diana (49s):
Hi Nicole How are you doing today?

Nicole (52s):
Great. I got up very early this morning and I’ve been working on my Dyer surname project stuff and it’s been so fun. What about you? What are you doing?

Diana (1m 1s):
Well, I am working on reading Roberta Estes’s book on Family Tree DNA, and I’ve been reading the section on YDNA, so we’ve both been doing a little bit with YDNA and just enjoying her explanations of all the different features that are on Family Tree DNA. I think so often when we click in there we just go straight to our matches to see if we have any surnames that match up with our tester’s surname. And it’s been fun to see her explanation of all the different things you can click on and what it means and how it can help you understand your results better.

Nicole (1m 37s):
Yeah, I’m excited to look at her book too. I just ordered it because sometimes I feel like there’s just so much to know and understand about YDNA and it just helps to have some resources. And I have David Vance’s book about YDNA and it’ll be nice to have Roberta’s as well.

Diana (1m 55s):
Right. I think that there’s a lot to learn and understand. There’s some really basic concepts that we teach, but if you’re working with a YDNA project, you’ve done the Big Y 700, you really wanna dig in and understand what you can do with that.

Nicole (2m 10s):
Yeah, absolutely. Well, like I mentioned, I’ve been writing some information for my Dyer surname project, and it’s been interesting. I’ve used the public results for the Dyer surname project for the last several years trying to figure things out. And I finally just got to the point where I was frustrated because a lot of people were confused about the earliest known ancestor and some subgroups were overlapping with the same person. So I finally just decided I would like to volunteer to help with this surname project because I want to help build all the lineages accurately and document them. And what you’re able to see on the public results page is pretty limited.

Nicole (2m 52s):
It only shows a kit number, no contact information, and what the person said their earliest known ancestor was, and then their YDNA markers STR markers. So in order to eliminate different Dyer families from consideration for candidates for my John Robert Dyer’s family, I needed more information. And since the research has been so challenging in Tennessee and there are a lot of other Dyers in Tennessee, I just thought, you know, it’d be great to be able to work on this and collaborate a little bit more with people. So I’m excited that I’m now a co-Admin of the project, and I decided to make a webpage on FamilyLocket that will just tell people about what I’m working on, and I can send that link to individuals when I ask them to take a YDNA test because I’m looking for DNA test takers from particular lines that I don’t think are represented yet, and that might be related to my Dyer line.

Nicole (3m 55s):
So that’s what I’ve been working on, and it’s been really enlightening and I’ve been really learning a lot about the history of surnames and when they began, and the Dyer surname, you know, beginning in several places all around the same time, around, you know, 1100 to 1200 AD and how in Ireland it was kind of a patron name based on people who were descendants of somebody with the name Duibhir, meaning dark. And then that turned into O Duibhir, or Dyer and o’Dwyer. And then in England it was this occupational surname coming from the person’s occupation where they were a dyer of cloth.

Nicole (4m 35s):
And so it’s just been so helpful to really think about how surnames came into existence really not too long ago. And a lot of the matches that people have in their YDNA are going back to a common ancestor prior to that time before surnames were established. And so they’re going to have matches to people with different surnames than their own, even if there’s no break in their YDNA line. And that can be confusing.

Diana (5m 1s):
Right. Well that was such a great example of using YDNA and the reason why you might want to learn more about it. So it’s so fun to be following your dire YDNA project.

Nicole (5m 11s):
Right. I needed to dive into this again before we start the study group because I’m hoping to continue working on it for the study group, but I had never wrapped up the rest of my DNA study group project and finished writing the report. So I’m getting all that wrapped up so I can get started on the next phase.

Diana (5m 29s):
Great. That’s a good feeling.

Nicole (5m 31s):
It is, especially when I’ve been putting it off all summer and I’m a little behind. It’s okay because it was worth it. I, I had to take my research trip to Oklahoma and work on my KDP for certification. So it was all for a good cause.

Diana (5m 44s):
And that’s a great thing about having all your research recorded in your research log, your timeline, you know, all the things you’ve done, you can just come right back to it and pick right up where you left off.

Nicole (5m 55s):
So true. Well, for announcements, our webinar for the month of October is on October 15th with Emma Lowe and Emma will be presenting Crossing the Pond: Tracing Dorothea’s Roots in 19th Century Pomerania. And this is gonna be a great study about Germany, immigration, evangelical Lutheran church records, German civil registration and more. So we hope you’ll join us. And if you haven’t registered for that, when you do register for $60, you get access to all of the monthly webinars, which include a research report with each lecture. And if you, if the lecture has already passed, you can watch the recording and for upcoming months, you can watch live on Zoom if you wish to, and you’ll receive lifetime access to those.

Nicole (6m 41s):
The next DNA study group is going to be starting in February of 2025 and registration will begin December. So start thinking now if you’d like to join us for that. And peer group leaders will receive complimentary registration. And so send us a report that includes DNA evidence if you’d like to join us as a peer group leader. If you’re not a member of our newsletter and you’d like to receive notifications of new blog posts and new podcast episodes and upcoming classes and coupons and things, then definitely sign up for our newsletter on our website at FamilyLocket.com/sign-up. And for upcoming conferences, we hope you’ll join us at the Texas State Genealogical Society Family History Conference.

Nicole (7m 24s):
It’s virtual on November 1st and second, and Diana will be presenting several lectures there. Well, for today’s topic, we are going to talk about research planning. And as Genealogists, we often get excited about looking at the records and searching and finding information, but taking the time to create a plan can make our research efforts much more productive. So today in this episode, we’re going to talk about the elements of effective research planning and we’ll talk about how artificial intelligence tools can help us. So first, let’s just talk about why research planning matters. It helps us stay focused on our research objective.

Nicole (8m 5s):
And like we were talking about, it helps you to be able to pick up where you left off. If your work is interrupted for a day or a week or even a couple months, it helps you to approach your problems systematically, like experts do. And often beginners will cover a lot of ground quickly by researching, researching, researching, but they miss really important records or methodologies that would help them because they didn’t take time to think about the FAN club and finding candidates for a hypothesis and different things where when you really study the known facts and make a plan, you realize you need to do well.

Nicole (8m 47s):
What are the elements of a research plan? Usually it includes an objective, a summary of the starting point information or known facts, a working hypothesis, the identified sources and the prioritized research strategy. So we’re going to talk about each of these elements in more detail using Diana’s case study to research Isabella Weatherford who lived in Texas in the mid 1800s. Well, first of all, the research objective, and this should clearly state what you aim to discover. In our first episode about this research project, we talked about using AI to write Isabella Weatherford, to write the objective for the Isabella Weatherford Project.

Nicole (9m 33s):
And the objective is to examine the economic and social conditions in Dallas County, Texas in the early 1870s and their influence on Isabella Weatherford’s life and marriage prospects. She was born in 1858 in Missouri, and first married John H. Carpenter in 1874 in Dallas County, Texas. Then she married Robert Sidney Royston in 1877 in Van Zant County, Texas, and she died on May 9th, 1942 in Tucumcari, New Mexico.

Diana (10m 5s):
All right, so let’s work on the second part of a research plan, which is coming up with a summary of known facts. So this section outlines what you already know about your research subject based on the previous research. So we talked previously in part two of the series about how I used AI to transcribe and organize a 57 page widow’s pension file for Isabella. And I had added all of those details to my Airtable timeline and then saved that timeline as a CSV file. And I uploaded that file to Claude 3.5 sonnet and prompted that AI model to create a summary of known facts for me in a table.

Diana (10m 53s):
So my Airtable timeline was pretty big, it had a lot of information, but the summary of known facts needs to be a little more simple. This is something we just look at to remind us of all the important things that we know. And usually I would take my timeline and I would create that manually, but I wanted to see if AI could help me. So my prompt was, you are an expert genealogist. I am researching Isabella Weatherford Royston attached is a detailed timeline for her life events from this timeline, extract names, dates, life events and places, and create a summary in the form of a table.

Diana (11m 34s):
The table should have four columns. And then I told it what the four columns should be, one for date and location, one for the person, one for specific details about the event and one for the source citation. Well, Claude did a great job creating the summary table and it was so easy for me to fact check. ’cause of course I wanted to make sure it was actually putting the right dates and places and names in, but I had the exact details in my Airtable timeline. So next I wanted the model to give me the data in a dot csv file And. so I did that. I got it imported into a spreadsheet and I, in thinking about it, I really like to just have a bulleted list for my project document because I take this summary of known facts and put it into my project document, which is not a spreadsheet, it’s a Google Doc.

Diana (12m 29s):
And so I wanted to have it in a bulleted list. And so then I just asked Claude to create the same information in that bulleted list with source citations and it did it perfectly. And then I could just take those bullets, copy ’em right into my project document. I did have to manually do the the citations into footnotes, but that didn’t take, you know, too long. And I suppose I could have used markdown for that and Writage, but I didn’t. I just did it manually. Nicole, would you have done that extra step?

Nicole (13m 0s):
Oh, definitely. Because it removes 10 extra steps later.

Diana (13m 5s):
Yeah, I guess I’m just not as comfortable doing that and so I didn’t, but…

Nicole (13m 8s):
Well, I don’t think you had Writage yet. You didn’t get it until you wrote the report, right?

Diana (13m 12s):
That’s true. Yeah, that’s true.

Nicole (13m 14s):
Well that’s fine. And you were just working with what you had and when you really wanted to try Writage, you got it and then now you know how to use it.

Diana (13m 23s):
That’s so true. You are right. Yeah, and this part of the process, I hadn’t gone that far, so yeah, but it still saved me a lot of time in creating all this for me really quickly. And I didn’t have to correct anything with any of the dates or events, which I was really happy about. But you know, that is because probably I uploaded the text and I said, this is what you’re using. you know, use the exact names and dates, places, so that might’ve helped.

Nicole (13m 52s):
Absolutely. And we know that’s one way to really get the models to not hallucinate is by providing the information we want them to use and then asking it to put it in a different format or do something with it and transform it in some way. And that’s actually one of my favorite things to do with it right now, especially while I’m working on a project actively. I find myself constantly wanting to take pasted information and reformat it in a different way for something. So copying and pasting a list of children into ChatGPT and say, format this as a bulleted list with all the names and dates and on one line per person. And so just taking it out of my Ancestry tree and getting it in that format so quickly, or taking a census extract from a transcription page on Ancestry and pasting that in to ChatGPT and saying, format this as a paragraph or however you want it.

Nicole (14m 47s):
And it’s just so fast. It is nice to not have to do it yourself,

Diana (14m 53s):
Right? I think it’s just always good to think about what we’re doing and could AI help us to do that a little bit faster or a little bit better?

Nicole (15m 3s):
Right. Well, I love how you were able to put the known facts into a CSV or into a bulleted list and looking at the bulleted list it made, it really was nice how it put the date first and then a little phrase about what it is, Isabella Weatherford born and then source, and then it has the citation right there. And that could have been just fine and you could have even done it as a table with the source and the right hand column and the information in the left. I’ve done that before for my known facts. And there’s a lot of ways to do it that don’t require you to use the special plugin that we’ve been talking about, which is called Writage. And for those of you who don’t have any clue what that is, it’s just a simple plugin for Microsoft Word that allows you to copy and paste text, text that’s been formatted with markdown in an AI model and paste it into Word and preserve the formatting.

Nicole (15m 58s):
So if you’ve tried pasting the formatting before and it just shows up as asterisk and other little marks and you didn’t know why, it’s because it’s using a simple formatting language called markdown for formatting text. And so we like that tool. It’s not expensive, I think it costs less than $30 one time fee to download it and use it.

Diana (16m 22s):
It’s just another little thing to learn about if you’re interested.

Nicole (16m 26s):
Right. And if you use a lot of copy and pasting from chatbots into Word, you’ll definitely find it useful. Alright, well the next step after the known facts, that summary of known facts is the working hypothesis. And this is an educated guess about what might have happened based on the known facts. It’s basically trying to answer your research question by taking a stab in the dark or an educated guess. I guess it’s not really a stab in the dark because an educated guess is more of a an estimate based on the clues you have. Well, AI tools can help us generate hypotheses and help us make our worth working hypothesis by analyzing the data we give it and also historical context.

Nicole (17m 15s):
So Diana, she was curious about comparing her hypothesis that she wrote with the one that AI would create. So let’s go over the one that she wrote. Isabella’s family moved from Springfield, Missouri where she was born on March 4th, 1858 to Dallas County, Texas. By September 3rd, 1860, the date of the census enumeration, the Civil War broke out the next year and Texas joined the Confederacy. This would’ve affected her family life with her older brother John possibly serving in the military. Reconstruction and the post civil war era would have also influenced life in Dallas County. The research will need to center on Isabella’s father Henderson Weatherford, as the records will be in his name.

Nicole (17m 59s):
The family has not been located on the 1870 census, so tax records could perhaps lend clarity to the family situation. There is no listing for Precinct six, so the broad Dallas County tax list could be searched. Henderson worked as a blacksmith in 1860 and perhaps continued the trade. Locating the Weatherford family in 1870 would be helpful. Isabella married John H. Carpenter on January 2nd, 1874 in Dallas County at the age of 15. He may have traveled to Dallas County on the railroad that arrived in 1872. He may have been a Confederate veteran from any southern state. Their daughter, Mary Clem Carpenter, was born October 3rd, 1875.

Nicole (18m 41s):
John either died or deserted Isabella by 1877 when she married Robert Royston. So that was a really good hypothesis that you wrote. Good job.

Diana (18m 50s):
Well thank you. It’s always fun to write those out.

Nicole (18m 53s):
Well it is good. It’s good to think about what we already know about what might have happened then. You provided Claude with your research objective and the summary of the known facts and prompted Claude to create a hypothesis using sound genealogical methodology. The response was fairly good and it included some research possibilities that hadn’t been considered yet. Some of the ideas that were particularly good were to research the economic conditions in Dallas County in the early 1870s. Look out for information about both of Isabella’s husbands to understand their backgrounds and circumstances, compare Isabella’s marriage patterns with those of other young women in similar circumstances.

Diana (19m 36s):
Well, it was fun to just see what Claude would come up with and I did notice that it gave me some new ideas and thoughts. So this is what I thought when I compared it, that my hypothesis, I was very specific with names, dates, and places. And I used my past research experience and knowledge about research in this era that generally means finding records for the men in a woman’s life. And I focused on the most likely scenario for the people involved, Isabella, her husband and her father. And I noticed that Claude provided new ideas like the economic conditions and gave broader searches such as looking for patterns in other marriages and divorces.

Diana (20m 20s):
And also it didn’t use any specifics, although they were right there in the summary of known facts. So that was just an interesting observation and I probably could have continued to ask it to be more specific or you know, give it some more prompting. But I actually really liked what it gave me because it made me think outside the box, which is good. Now let’s have a word from our Sponsor. Today’s episode is sponsored by Newspapers.com. Break down genealogy brick walls with a subscription to the largest online newspaper archive. Did you know Newspapers.com has over 1 billion pages of digitized newspapers dating back to 1690. Their growing collection includes papers from the US, UK, Canada, Australia, and beyond.

Diana (21m 3s):
Discover birth and marriage announcements, obituaries and everyday stories about your ancestors in seconds. Newspapers.com can help you fill in the gaps between vital records and reveal details about your ancestors’ lives that you can’t find anywhere else. They’re easy to use search feature lets you filter your results by date, locations specific paper and more. When you find something interesting, Newspapers.com makes it a snap to share it with family and friends. You can even save it directly to your Ancestry tree. Come explore 1 billion pages and make infinite discoveries today on Newspapers.com. Use promo code FamilyLocket for 20% discount on your subscription

Nicole (21m 42s):
Thank you. The next step of a research plan is to identify sources and this is where you brainstorm and list all potential sources that might contain information relevant to your objective. That might help you answer your research question. So you could make a list of different things to look at. Census records, marriage tax lists, newspapers and more. AI tools can help with this step by suggesting sources based on your locality guide if you upload that and based on your research objective. We did talk about using AI to help with locality research in a previous episode and it’s really good at that. That’s a helpful way to use AI.

Nicole (22m 24s):
One of the really great tools that we like is using Claude, which is a chat bot by Anthropic. And Claude has a feature that allows you to set up a project and if you have a pro account where you pay about $20 a month, you can create projects. And basically what you do is create a knowledge base that Claude will draw on for many of your chat conversations. And anytime that you want to chat about that project, you just select the project. So you can copy and paste information into the knowledge base, you can upload documents and there is a memory limit, but it’s fairly large. So you can easily upload your whole locality guide and several previous projects that you’ve done with the research report.

Nicole (23m 9s):
So if you have, you know, four research reports that are 20 pages each, so we’ll be fine to upload them. Once you’ve done this for a research planning step, which you might want to upload, like your objective, your background information and summary of no facts and your hypothesis, then you can save that as a project and then you can chat with Claude about that project. And so here’s an example of how Diana did that. She prompted Claude to create a list of sources that could be used in the research. The results were somewhat helpful. There were some really good sources that Diana agreed with that she would do.

Nicole (23m 49s):
Those included the 1870 federal census, Dallas County tax records, Dallas County probate records, deed records, newspapers from Dallas County Civil War records, a history of Dallas County and another history of Dallas County. So it did a good job and it seems like it really drew upon your locality guide to give complete information about these sources and didn’t just list them, but actually gave their full title and some ideas of what you would search them for,

Diana (24m 22s):
Right? I really appreciated that. I thought that it gave some good, insightful thoughts about how to attack this problem. So after you know we have this long list of possible sources, then we need to do a priority. And that’s the very final step in research planning. And that’s when you wanna organize your sources into a logical order. And so we always think about things like how easy is it to find the records and how likely that records will hold the relevant information and the specific time period we’re researching in. So AI tools can also help prioritize these based on these factors and your specific research goals.

Diana (25m 4s):
So for priority, Claude, when I asked that to do this gave the following priority: first do the 1870 census, then tax records, newspapers, deed records, and city directories. So that one’s interesting. And I decided to also see what ChatGPT would do with this very same bit of information. So I’ve been talking about how I used Claude quite a bit in this research planning process, so I decided to use the principle of prompt chaining for ChatGPT. so I uploaded basically the same information. I gave it the timeline, the objective, a hypothesis and historical information from the timeline and asked it to give me a hypothesis.

Diana (25m 52s):
And then I asked ChatGPT to do a research strategy based on the locality guide. I then uploaded and then I asked for a prioritized research strategy. so I basically went through all the same steps with a different AI model just to see what that model would give me. And I actually really liked that quite well. It was different than what Claude gave me, but it was pretty similar. So ChatGPT suggested first to do the county histories and then the 1870 census, then newspapers, then land and property records and probate. So of course I also created my own research priority according to my experience researching in Texas, which I have done for years.

Diana (26m 38s):
And I also like to bring up the point always that our intuition, our gut feeling about what we should, should research first based on all of those things is really good. So I thought it would be fun to compare the priority that we each Claude, ChatGPT, and myself gave. So I also put the history first, which agreed with ChatGPT. And then I decided that newspapers would be second, which both the models had put as third. And then from my third part I put the census research for the 1870 census, which Claude gave first and ChatGPT put second.

Diana (27m 20s):
And as a listener you might be thinking, well why didn’t you do the census research first? Well I had looked high and low for that 1870 census and so I kind of put in there as an afterthought just to do it and say, okay, I searched it so I could have it as a negative because I’ve looked for so many years for that. But it will be fun to reveal what what I found when I actually did that part of the process in future episodes. And then fourth I put tax research, which Claude put second and ChatGPT didn’t list at all. And then fifth I put migration to Dallas County, Texas, which neither AI model had in their list at all. So you know, it was was kind of fun to do the comparison and see, you know, how my ideas, my experience came up basically with many of the same things but some different things,

Nicole (28m 8s):
Right? And something unique was that you wanted to look into how they got to Dallas and thinking about the railroads and all that. So, and I think that’s an important thing that we don’t leave at all up to the AI, that we think of our own ideas and what we would do because the artificial intelligence is sometimes just that, just artificial and it’s just giving ideas based on what it’s read in the past. But we have to come up with those new unique creative ideas that only a human can think of. Well I know that next time when we talk about this project, you’re gonna give us some feedback on how the research plan went, but could you give us just a hint of was it super important the order that you did all these things in or did you just do all of them?

Nicole (28m 51s):
And so the order wasn’t as important.

Diana (28m 54s):
I did end up doing most of these and I found something in my 1870 census search that pretty much changed my whole thought pattern of this. So look, looking back, ChatGPT, or no, it was Claude that said it should do that first. It might’ve been good to have done that first. So it was kind of funny. It’s always interesting to look back at your plan and think, oh, I should have done that first. But you know, it is what it is. It’s okay.

Nicole (29m 23s):
Well I think as long as it’s included in the top five and you got to do it, it’s not as important as what order you do it in unless you spend a ton of time doing numbers one, two, and before you die.

Diana (29m 36s):
That’s right.

Nicole (29m 36s):
That’s why I tend to put easy and fast ones first because I don’t wanna spend a ton of time if there’s something easy to find.

Diana (29m 43s):
Right. And I kind of envisioned the census search taking a long time, like really paging through every page, you know, that type of thing. Yeah. Which is why I put a little bit later.

Nicole (29m 54s):
So good job. Well thanks for sharing your thoughts about all of that.

Diana (29m 59s):
Of course. Well, you know, and just to, to wrap this up, hopefully everyone listening has had a good refresher on research planning and how valuable it can be for any type of research problem. And AI was a fun addition to this. It gave me some new ideas, new perspective, and I think if we’ve been using the same strategy for years upon years, you know, sometimes we get in a rut with our research and we just use the same four or five sources. I think it’s good to give us some new ideas. so I would encourage everyone to give it a try and see what you can find.

Nicole (30m 36s):
Alright, well thanks for sharing that and I hope you all have a great week and we’ll talk to you again next week. Bye.

Diana (30m 43s):
Alright, bye-Bye. Thanks everyone.

Nicole (30m 43s):
Thank you for listening. We hope that something you heard today will help you make progress in your research. If you want to learn more, purchase our books, Research Like a Pro and Research Like a Pro with DNA on Amazon.com and other booksellers. You can also register for our online courses or study groups of the same names. Learn more at FamilyLocket.com/services. To share your progress and ask questions, join our private Facebook group by sending us your book receipt or joining our courses to get updates in your email inbox each Monday, subscribe to our newsletter at FamilyLocket.com/newsletter. Please subscribe, rate and review our podcast. We read each review and are so thankful for them. We hope you’ll start now to Research Like a Pro.