BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//conferences.acspri.org.au//2026//speaker//FDEFGD
BEGIN:VTIMEZONE
TZID:AEST
BEGIN:STANDARD
DTSTART:20000326T040000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3;UNTIL=20050326T170000Z
TZNAME:AEST
TZOFFSETFROM:+1100
TZOFFSETTO:+1000
END:STANDARD
BEGIN:STANDARD
DTSTART:20060402T040000
RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=4;UNTIL=20060401T170000Z
TZNAME:AEST
TZOFFSETFROM:+1100
TZOFFSETTO:+1000
END:STANDARD
BEGIN:STANDARD
DTSTART:20070325T040000
RRULE:FREQ=YEARLY;BYDAY=4SU;BYMONTH=3;UNTIL=20070324T170000Z
TZNAME:AEST
TZOFFSETFROM:+1100
TZOFFSETTO:+1000
END:STANDARD
BEGIN:STANDARD
DTSTART:20080406T040000
RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=4
TZNAME:AEST
TZOFFSETFROM:+1100
TZOFFSETTO:+1000
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20000827T030000
RRULE:FREQ=YEARLY;BYDAY=4SU;BYMONTH=8;UNTIL=20000826T170000Z
TZNAME:AEDT
TZOFFSETFROM:+1000
TZOFFSETTO:+1100
END:DAYLIGHT
BEGIN:DAYLIGHT
DTSTART:20011028T030000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10;UNTIL=20071027T170000Z
TZNAME:AEDT
TZOFFSETFROM:+1000
TZOFFSETTO:+1100
END:DAYLIGHT
BEGIN:DAYLIGHT
DTSTART:20081005T030000
RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=10
TZNAME:AEDT
TZOFFSETFROM:+1000
TZOFFSETTO:+1100
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
UID:pretalx-2026-FTTU3E@conferences.acspri.org.au
DTSTART;TZID=AEST:20261126T110000
DTEND;TZID=AEST:20261126T111500
DESCRIPTION:Coding qualitative open-text data is a central concern across t
 he social sciences. Survey researchers develop and apply codebooks to part
 icipants' open-text responses in order to derive structured data. More rec
 ent digital research methods scrape unstructured content and likewise iter
 ate through text\, images\, and other media to describe a corpus in a stru
 ctured manner. Numerous research papers have explored large language model
 s (LLMs) as a means of automating this coding process. These studies sugge
 st that reliable\, complete agreement with 'gold-standard' codings (i.e.\,
  two independent experts code responses and reconcile their results) is un
 likely to be attainable\, even with frontier models. However\, an LLM code
 r that achieves a level of inter-coder agreement with a human coder compar
 able to that achieved between two human coders may be attainable. Given th
 at this level of performance appears possible with frontier models\, a nat
 ural next step is to determine whether comparable performance can be achie
 ved using smaller models.\n\nThis research presents a system that uses dis
 tilled and quantized language models small enough to run on locally deploy
 ed\, GPU-accelerated commodity hardware. As a result\, the approach can be
  effectively utilized on devices as modest as a laptop computer.\n\nWe pre
 sent a locally running automated open-text coding agent built on a small o
 pen-weights model. This approach offers several advantages. First\, it imp
 roves replicability. Frontier models do not guarantee that identical outpu
 ts will be returned indefinitely\, whereas open-weights models can be conf
 igured to consistently produce the same results. Second\, it enhances data
  privacy. Researchers often need to engineer systems that ensure sensitive
  data\, including participants' open-text responses that may contain perso
 nally identifiable information\, never leaves their control or is transmit
 ted to third-party servers. A locally deployable system makes secure on-pr
 emises processing straightforward.\n\nThird\, open-weights models provide 
 greater flexibility for customization. In the current implementation\, we 
 use a simple classifier to estimate the probabilities of binary "Yes/No" o
 utputs per code definition and derive one-versus-one code probabilities. T
 his enables classifications to be ranked by confidence and allows threshol
 ding techniques to identify genuine "none of the above" responses. We can 
 also implement guided reasoning workflows that improve the quality of codi
 ng justifications and evidence-based grounding.\n\nFinally\, deploying sma
 ller models on hardware of the researcher's choosing allows per-token cost
 s and energy consumption to be reduced. Researchers can also invest in gre
 ater GPU acceleration or parallelize inference across multiple machines to
  process large corpora quickly. This enables rapid iteration of codebook d
 esign and supports efficient large-scale qualitative analysis.
DTSTAMP:20261102T131145Z
LOCATION:Sutherland Room
SUMMARY:Qualitative Analysis with Small\, Open-Weight Language Models: An A
 uditable\, Local\, Self-Hosted Workbench for Computer-Assisted Qualitative
  Analysis - John Collins
URL:https://conferences.acspri.org.au/2026/talk/FTTU3E/
END:VEVENT
END:VCALENDAR