[HN Gopher] Show HN: RoundtableJS - Open-source programmatic sur...
___________________________________________________________________
Show HN: RoundtableJS - Open-source programmatic survey library
Hi HN, we're Matt and Mayank of Roundtable. We're launching
RoundtableJS, an open-source library
(https://github.com/roundtableAI/roundtable-js) to program research
surveys at scale. Academic and market researchers want a DIY
platform for survey research. Most existing tools are no-code, and
these researchers want programmatic software that handles advanced
branching logic and customization. We know this because we spent
five years using open-source survey software for our academic work.
One of the key insights of RoundtableJS is to leverage JavaScript's
async/await functionality for managing survey logic. In other
libraries survey logic is determined via callbacks that trigger
when a page is submitted (i.e., if the user answers "Yes" to Q2,
show page 3), which quickly becomes a nightmare to manage and
debug. Our library lets developers build the timeline in an async
function which means the logic flows intuitively from top to
bottom. However, this fully programmable logic can be inaccessible
to non-developers, and so our cloud offering
(surveys.roundtable.ai) lets users program surveys using natural
language by storing the library into context of an LLM (which acts
as a library-specific copilot, demo link here:
https://www.loom.com/share/7e7f30aa96244542a62d6c6ed125f2d5?...).
We think that being open-source will give us an edge in integrating
LLMs into our platform. We designed RoundtableJS with LLM usage in
mind, i.e. making it fully customizable and general-purpose. The
alternative - building a library with a more constrained focus -
would be more intuitive for unassisted programmers, but AI
assistance makes it really easy to customize any part of the
survey. Already we think this approach is easier and faster than
no-code alternatives. We left academia last year and have been
building products in the survey space since. Our first product was
an AI-powered survey simulator
(https://news.ycombinator.com/item?id=36865625). People loved
playing with it, but it was difficult to monetize (as many HN
comments predicted). We abandoned this, talked to customers, and
found that data quality is a huge problem in market research (e.g.
https://www.kantar.com/company-news/kantar-partners-with-rea...).
We released an API that analyzes open-end responses for low quality
and fraudulent typing behavior. This helps firms save time cleaning
data, but it is a limited market. We plan to monetize through our
cloud offering, which currently offers instant link deployment, an
interactive code editor and a fraud detection suite. We plan to
introduce more features here in the analytics, design, and
productivity space. Surveys are ubiquitous, but they are often
bland and generic. By allowing people to fully customize them for
today's distribution channels, we want to make it easy to design
better, more engaging surveys. We just launched. Where can we
improve?
Author : timshell
Score : 32 points
Date : 2024-08-08 16:13 UTC (6 hours ago)
(HTM) web link (github.com)
(TXT) w3m dump (github.com)
| timshell wrote:
| Cloud editor open access for today at surveys.roundtable.ai.
|
| Username: show-hn@news.ycombinator.com PW: Hackernews2024!
| jeanlucas wrote:
| Great. We really need more open source tools for surveying and
| forms.
| swyx wrote:
| lovely, congrats on launching! i guess at the limit this is a
| qualtrics competitor? and before that i guess google forms. can i
| use this as a strict google forms replacement? - dump data
| straight into a google spreadsheet?
|
| i liked
| https://roundtable.ai/survey/9593255a3aa7231051e75e578747eae...
| as a multistep option
|
| good luck - i think its an important problem but idk if its
| venture scale yet. you might explore RLHF/AI data annotation as a
| usecase.
|
| i'd be most concerned about botting and filtering out low effort
| surveys - the latter i imagine is a big academic concern. if you
| came out with a singular innovation there you could probably
| build a company around it.
___________________________________________________________________
(page generated 2024-08-08 23:01 UTC)