Research

What happens when students discuss with CAiSEY before class

CAiSEY was built by career researchers. We develop it on rigorous evidence and publish original findings on how new learning modalities shape educational, cognitive, and social outcomes. Every study we share is IRB-approved and pre-registered where applicable, with replication packages and data transparency statements available on request.

Randomized field experiment

Two CAiSEY conversations, more class participation all term

~800 MBA students in sections of the same course, each randomly assigned two of ten sessions to prepare with a CAiSEY conversation. Instructors logged every voluntary contribution in class.

After a second CAiSEY conversation, students who completed it contributed 31% more in class discussion for the rest of the term.

Request working paper
+31%class participation after the second conversation
+20%+40%+60%Average +31%23456789101112+49%1st conversation2nd conversationClass session
Voluntary class contributions after the second CAiSEY conversation, compared with the same student having only the first one, by class session. Average +31% (p = 0.039).

Randomized field experiment

Written case work becomes more original and less AI-formulaic

Same experiment, ~7,000 short written assignments submitted by ~800 MBA students. Each was submitted minutes after the assigned CAiSEY conversation, so the comparison is the same student's writing with and without a conversation first.

After a conversation, students wrote with a wider vocabulary, cited more evidence, and produced text that read as measurably less AI-generated.

Request working paper
−0.25 SDAI-formulaic writing after a conversation
LessMoreVocabulary range+0.24 SDLonger words+0.19 SDRange of ideas+0.17 SDEvidence cited+0.14 SDAI-formulaic writing−0.25 SD
Change in the written assignment after a CAiSEY conversation, in standard deviations. Positive bars are increases, negative bars are decreases.

Published preprint

Talking to CAiSEY opens up a wider range of ideas than typing

~1,000 voice- and text-based conversations with CAiSEY submitted by 200+ MBA and EMBA students in a business school elective class. Students chose voice or text for each case; the reading, the prompt, and CAiSEY's opposing stance were identical either way.

Talking opened up a wider range of ideas than typing (semantic diversity) even after adjusting for the extra words spoken, signaling that students achieve greater creativity and cognitive flexibility through voice-based AI conversations.

Neelam Modi Jain and Dan J. Wang, Columbia Business School

Read the paper on arXiv
+31%wider range of ideas (semantic diversity) when talking vs. typing
TypingTalkingWords per conversation≈250≈500Unique words≈110≈190Messages sent≈7≈9Semantic diversity4.55.9
Typical conversation by modality. Semantic diversity is an index of how many distinct ideas a student's arguments span; higher means broader, more flexible reasoning.

Bring CAiSEY into your classroom

Interested in bringing CAiSEY into your course or institution? Let’s talk about research, pilots, and classroom use cases.

Get in touch