Research with Dataspheres AI - in 10 Minutes | DATASPHERES AI - Dataspheres AI

From Field Recording to Findings DATASPHERES AI From Field Recordings to Findings Run a complete study on one platform. [email protected] nicole.bao...

From Field Recording to Findings DATASPHERES AI From Field Recordings to Findings Run a complete study on one platform. [email protected] [email protected] 0:00-0:40 - Welcome. One sentence on who we are: 'We build a research platform for communities and the people who study with them.' Then straight into the story - no agenda slide, no bio. The deck is 10 minutes: scene (1), cost (1.5), the turn + the course (1), six modules (5.5), close (0.5). The course IS the talk - they leave with the async version of everything shown. It's 11 p.m. Industry Pain Points Zoom Recordings, Google Drive, Audio/Video on your local computer... Otter Transcriptions (...another Subscription to manage) Codes live in NVivo - on the one lab machine that has the license. Consent & Surveys buried in Qualtrics. The best quotes are somewhere in a Word doc. And the findings deck is due Friday. 0:40-1:40 - Read the inventory slowly; every line is a real tool the room pays for and fights with. The specificity IS the persuasion - generic pain slides bounce off, but 'codebook_v7_FINAL_revised.xlsx' gets a laugh because everyone has that file. Land the gold line, pause: 'Sound familiar?' Wait for the nods. Fragmentation has a price Fragmentation has a price Traceability "Which voice does this claim come from?" Time Hours spent re-finding things you already found. Trust A reviewer's question - or an IRB audit - you can't walk back to the data. Money A QDA seat can run $800-1,400 a year. Transcription is another subscription. ...multiplied by each artifact & each analysis... 1:40-3:00 - Agitate honestly: these are costs the room already pays. Traceability, time, trust - then money: 'a QDA seat alone can run eight hundred to fourteen hundred dollars a year, per person, and transcription is a separate subscription on top.' Land trust hardest - 'every claim you can't trace is a claim someone else gets to question.' That earns the turn. A free course, six modules What if the whole study lived in one place? We packaged the whole arc into a free, self-paced course - six modules. One datasphere holds the study - recordings, transcripts, codes, datasets, surveys, reports - with dashboards in sync in real time. This is the directory of one: Sample Datasphere - AI in Academia 🔗 ↗️ 3:00-4:00 - The turn, and the plan. 'What if the whole study lived in one place? We built that - and we packaged the whole workflow into a free course you can take async.' Point at the screenshot: 'This is a datasphere.' Then: 'Here are its six modules in six slides.' Module 1 - Your Research Workspace MODULE 1 OF 6 - YOUR RESEARCH WORKSPACE A home for the whole study - and you hold the keys The datasphere as a research container: roles and access controls scoped to your study, privacy walls around participant spaces, audit trails on every AI-assisted step. In the sample study: a public workspace anyone can inspect - dataspheres.ai/pages/ai-in-academia 4:00-5:00 - Module 1 sets the container AND answers the IRB question up front: 'everything you're about to see runs inside access controls you configure, and the platform documents its own security architecture so you can cite it in a protocol.' Control language from the first minute. Module 2 - Collecting Data MODULE 2 OF 6 - COLLECTING DATA Everything lands in one place Interviews, video, documents, images, participant journals, and continuous surveys - uploaded from any device, straight into the study workspace. 5:00-6:00 - Module 2. 'Multimodal, at scale - a phone at a field site and a laptop at the office feed the same workspace.' Continuous surveys mean intake keeps flowing while you analyze. Module 3 - Transcription for Real Fieldwork MODULE 3 OF 6 - TRANSCRIPTION FOR REAL FIELDWORK Voices become text - speakers kept apart Speaker-diarized transcription 10 speaker audio sample The transcript is a first draft you review & edit. 6:00-7:00 - Module 3. 'Diarized means every line knows who said it.' Multilingual matters to this room - fieldwork rarely happens in English only. Module 4 - Qualitative Coding with an Audit Trail MODULE 4 OF 6 - QUALITATIVE CODING WITH AN AUDIT TRAIL Your codebook, your call Thematic analysis with audit trails and manual controls at every step - AI assists as a second coder, never the author. You accept, reject, or rewrite every application. 7:00-8:00 - Module 4 - the slide this audience is most skeptical about, so give them the receipts: 'In our sample study we double-coded 53 segments with two blind coders - kappa 0.89 on theme, 0.91 on stance. And the disagreements caught the original coder drifting: statistics inside critical arguments had been coded critical; the blind pass read them as balanced, the codebook was amended, and the adjudication is published. AI assists as a second coder - never the author - and we practice that on ourselves.' Control language backed by a published reliability page. Module 5 - Datasets, Demographics and Live Visualization MODULE 5 OF 6 - DATASETS, DE