So I’ve been reading a lot about the usefulness of ChatGPT-type AI to help research and identify and compare schools to help build a list, or even to help plan routes to visit schools for a multi-day college trip, that kind of thing.
I know there are free ones and paid versions; I’d be willing to try out a paid version if it seemed like a reasonable cost. I also know there are lots of different AIs (Claude, ChatGPT, Perplexity). Does anyone have experience using any of these platforms? Any particular recommendations? I’m just dipping my toes in but I’m hoping it might be useful as we are starting out. DC and I are both feeling a bit overwhelmed. |
Search on here.
Kollegio was mentioned along with others. |
I only had experience with ChatGPT, paid version. BE SURE TO TURN OFF THE DATA SHARING!!!
Okay, once you turn off the data sharing, you can ask the bot almost anything. DC uploaded everything to the bot, GPA, score, course rigor, high school profile, past matriculation, just any data you can gather! Essentially you are building a model specifically tailored for your DC. You can use the bot for give DC's viability score with regard to each individual college. Switching between ED and RD options. Mapping out where DC stands in terms of their competitiveness. |
Your school matters the most. All the other data are remotely relevant. By building a model tailored for your school and specifically tailored for your DC, the feedback you got becomes more meaningful because they are provided in the context of your school. |
I uploaded our school stuff too. Into paid. No data sharing. Very helpful. |
I am trying to do this now. Sort of a newbie using AI. I have Claude paid. So I upload all the info listed above, including school profile, etc. Does AI store this info into a file I can retrieve the next time I log on? How do I do that, so I don’t need to reload it every time I use it. Sorry if this is a dumb question. |
Yes. You don’t have to reload. But worth keeping in mind that the LLM looks for patterns, essentially, based on its training set, which is the internet. So while this is a massive simplification, here’s an example: if people on DCUM and other web sites always post that kids who are interested in political science should go to William and Mary, the LLM will say “Because your child is interested in poli sci, I recommend William &Mary for its programs in…”. The LLM doesn’t have access to some greater or deeper knowledge, or actual experiences (beyond those posted online), or admissions files. It’s giving you the conventional wisdom essentially, the patterns that most closely match what you’re asking. It’s not evaluating in the human sense. It’s not that it can’t be useful, but it’s good to understand what it’s drawing on. |
Amending this to say: if you feed it data from scattergrams, naviance, etc, then yes, it can use that data for example to give a percentage chance. You just want to think about what it is using to generate results and truly validate those results. |
Use every AI, cross reference - Claude, chatGPT |
Save as a “project” on Claude. And title it. |
At our private school, they also circulate this at the end of each year of where kids were admitted to separate from Navion. So it’s like a chart. It may not be complete, but it’s a good additional data source. I also inputted anything I knew about legacies into a class when it was Assessing the quality of ex missions and to see where there might be opportunities. It had some good analysis and year to year trends. The more specific data it has the better it is. The more specific your prompts are the better analysis will be. Continue to refine the prompts over and over and over again and ask it really hard questions and question the assumptions it is making and even some of its results. Ask it to verify With, “are you sure XYZ” - sometimes the defensive analysis is really good. And sometimes it’s discovered it was “hallucinating”. |
ChatGPT et al draw upon whatever they find on the internet. . .and make up information when they run out of material to source. Also, if I were you, I wouldn't rely upon ChatGPT to map out routes for a multi-school visit. Honestly, if you and DC are feeling overwhelmed, what you should do is the following: 1. DC should make a list of the qualities s/he/they ideally would want in a school. 2. Assuming that DC is a reasonably competitive student, you and s/he/they should go to the library to check out the written descriptions of schools in Fiske's Guide to Colleges and/or check out the lists and descriptions here: https://www. princetonreview.com/college-rankings/best-colleges. The Princeton Review also has this platform where you can input your stats, interests, etc. and it will give you a list of possible college options based upon that data: https://inquiry.princetonreview.com/ugrad/studentslikeyou/?destid=ugrad_web_collegehub-tile_grpf_findmatched 3. DC should then consider his/her/their competitiveness as a student with the student profiles of the schools s/he/they like best. 4. After factoring in considerations such as cost (including travel between school and home), access to health resources (are good hospitals nearby? What reproductive health services will be available?), safety (and not just the more obvious stats but things like whether a given campus/state is open carry), etc., figure out where the schools of interest are in relation to each other and plan trips accordingly. |
Keep in mind that gen AI is just a predictive model.
It does crawl the web for info, but there's no telling if it's outdated and it doesn't really have the ability to discern credibility. |
Any school specific data would be highly useful. I also input college acceptance rates specific for our high school into gpt, some of those differ wildly from the general acceptance rates.
Other information include matriculation majors. You can analyze these majors either to find a trend or screen out popular majors. |
Oops. Thanks for the tip. What if you don’t? I just started using it, so haven’t shared much. |