Key Topics: Compliance Frameworks, Corporate Governance, Strategic Planning

CSteps current policy is to avoid the direct use of generative AI but to explore potential AI avenues that may yield rapid benefits in promoting climate solutions to the level that they outweigh AI's current environmental costs, such as massive water and electricity use, and that transparently provide data that can be and is trusted by the users.

Students will help Climate Steps pressure-test its current “no AI” posture by analyzing the policy, staffing, ethical, and reputational tradeoffs of adopting AI tools in an environmental nonprofit context for help in solving climate crises.  Environmental costs of AI and mega data centers include their rapid builds, often in rural areas and without transparency, where the grid, the communities, the non-coal power, and/or the water sources cannot support them.  One AI/data center build uses (soon "used") massive power generators without permits, polluting the air for people downwind, which has created a bad reputation for AI/data center builds in general.  The re-opening of coal plants for data centers is another threat for climate change and thus to climate-engaged users.  Thus AI use has has not only its current data-quality trust issues but severe environmental-cost issues for people connected to the environment, such as Climate Steps users.
 
However, AI-, data-center-use may be justified, as AIs do and could further help model the impact of climate change [when quality data are provided], and can help model climate solutions -- at a time when climate change is a critical issue to fix within as few years as possible. 
 
This work will focus on defining decision criteria for when AI use is justified for an environmental nonprofit, focusing on mission benefit thresholds, environmental cost accounting, data/information quality, transparency expectations, the risk of user backlash, and safeguards.   

This project will include significant research on the current and future environmental and data-quality levels of AIs and data centers (and the difference between  AIs and data centers).   

It will also need research on the current public concepts of the impacts of AIs, on how information about new technology (good or bad) is best shared with the general public, and on how stakeholders in Climate Steps (see other project) may react.   Note, although Climate Steps is strongly international, the focus here is on North American-based audiences.

Also note, that although this statement covers "AI" in general, Climate Steps is looking at potentially different uses of AI, and some uses of different AIs may be better in environmental and trust cost:benefit ratios than others. This can be tricky research/to summarize, as there are many different uses for AI already.  Climate Steps has provided examples of AI use it is considering in an attached section of a grant application.  How you decide to handle this will be part of the recommended policy report.

This project will then involve the construction of decision-rubrics, including an overarching one, re:
a) mission benefit thresholds
b) environmental cost accounting,
c) data/information quality, (Climate Steps is science-based; it's goal is to promote truly effective climate solutions.)
d) transparency expectations by the public (Climate Steps will always be transparent on AI/data-center use), 
e) the risk of user backlash, and 
f) safeguards. (for you to define.)

Climate Steps has attached it's Principle.doc, a section of a recent grant application that specifically answers a question to a funder on why Climate Steps will survive AI, and several recent articles on the ethical use of AI in science/environmental conservation. 

Are there any benefits of AI in creating climate action that truly outweigh the environmental costs?  

 Deliverables will include a report on costs:benefits of AIs for an environmental nonprofit, recommended AI policy (what’s permitted vs. prohibited), a decision rubric for evaluating future AI use cases, two case studies, and a communications-ready narrative explaining the policy to stakeholders and users.