Further Progress Toward Automating Functional Analysis Interpretation.
A free script now reads FA graphs almost as well as you do.
01Research in Context
What this study did
Adams et al. (2024) built a computer script that reads functional-analysis graphs.
They asked experts to look at the same graphs.
Then they checked how often the script and the experts agreed.
What they found
The new script matched expert eyes 89% of the time.
That is up from 81% in the last version.
The jump means fewer graphs need a second human check.
How this fits with other research
Lanovaz et al. (2021) already showed that machine learning can beat visual inspection on fake AB graphs.
E et al. now move that idea to real FA graphs and hit 89% agreement.
Jessel et al. (2020) warned that FA methods are still not standard after 50 years.
The new script does not fix the lack of standards, but it gives teams one consistent reader.
Coffey et al. (2020) say IISCA is a faster FA format; the script could read those graphs too.
Why it matters
You can let the script screen every FA graph before your weekly meeting.
It flags the clear cases so you only debate the tricky ones.
That saves billable hours and gets treatment started sooner.
Want CEUs on This Topic?
The ABA Clubhouse has 60+ free CEUs — live every Wednesday. Ethics, supervision & clinical topics.
Join Free →Run the script on last month’s FA graphs and see which ones it calls correctly.
02At a glance
03Original abstract
It is considered best practice to conduct a functional analysis and visually inspect data collected to determine the function of problem behavior, which then informs the intervention approaches applied. Visual inspection has been described as a "subjective" process that may be affected by factors unrelated to the data. Structured decision-making guidelines have been established to address some of these shortcomings. The current paper is a follow-up to earlier work describing positive outcomes related to the viability of a decision support system based on structured criteria from Roane et al. Here, we demonstrate important improvements in a computer script's interpretation of functional analysis data, including improvement in agreement between the updated computer script version and experienced human raters (89%) compared to our original agreement outcomes (81%). This paper further supports the use of decision support systems for functional analysis interpretation.
Behavior modification, 2024 · doi:10.1177/01454455231195825