About Me
Mike Mahoney
PhD Student at SUNY-ESF and ESIP Community Fellow
Focusing on visualization, VR and ML as ways to think about large-scale systems
These slides: mm218.dev/esip2022
New York State 2019 CLCPA:
85% reduction in GHG emissions by 2050
Net zero emissions statewide by 2050
Lands & forests make up the difference
Need a way to monitor sequestration
Figure from Shirer, R. and Zimmerman, C. (2010) Forest Regeneration in New York State. The Nature Conservancy.
Modeling approach:
- Ensemble modeling approach
- Ensembling improves predictive accuracy, lets us try many models but present a single prediction
The press will have you believe that machine learning can reliably predict whether you’re gay, whether you’ll develop psychosis, whether you’ll have a heart attack and whether you’re a criminal—as well as other ambitious predictions such as when you’ll die and whether your unpublished book will be a bestseller. It’s all a lie.
Here’s how the lie works. Researchers report high “accuracy,” but then later reveal—buried within the details of a technical paper—that they were actually misusing the word “accuracy” to mean another measure of performance related to accuracy but in actuality not nearly as impressive.
This is the mainstream opinion of AI and machine learning.
How can we build back lost trust when developing AI systems that might impact people’s lives?
Report accuracy spatially
![]()
Report accuracy across scales
| Scale |
RMSE |
MBE |
R2 |
| Plot-pixel comparison |
39.60 |
1.88 |
0.76 |
| 8,660 ha hex |
33.27 |
3.56 |
0.76 |
| 78,100 ha hex |
23.38 |
1.64 |
0.80 |
| 216,500 ha hex |
21.17 |
0.75 |
0.81 |
Report applicability domains
Remaining Challenges
- How can we tailor our assessments to be the most useful for end users?
- How can we avoid overwhelming people with charts and tables when they just want a map?
