Reflections
Each week during the forked lessons, you will submit one paper reflection based on one of that week's required readings. A good reflection goes beyond summarizing the paper—it should focus on your own critical analysis. See the rubric below for detailed expectations.
If you are looking for guidance on how to read and engage with research papers, the following resources may be helpful:
Rubric
Length: Up to 1 page | Frequency: 1 reflection per week
Assessment format: Complete/Incomplete
Summary (1–2 sentences)
- Give a brief, accurate overview of the paper's main contribution
- Establish the core claim or result for the reader
Context & Problem Framing
- How did the paper frame the underlying problem to be solved?
- Articulate clearly the gap the authors are addressing and why it matters
Methodology & Results
- Describe the method used by the authors accurately at a high level
- Reflect on the results: what do they demonstrate, and are they sufficient to support the claims?
Critique
- What aspects are you convinced by and why? What aspects are you less convinced by and why?
- Consider: applicability, soundness, key stakeholders, responsible AI considerations (e.g. bias and fairness, trustworthiness, robustness, accountability, accessibility, equity, dual use, cost)
- What would you have changed in the problem framing, methodology, or evaluation?
Extensions
- Propose at least one non-trivial extension that reflects deeply on the underlying problem
- Extensions should go beyond simple modifications (e.g., swapping in a new ML architecture) and may include new problem contexts, methodological directions, or application domains
Paper Presentations
Each group will have 15 minutes to present, with up to 25 minutes total allocated per group to allow time for questions and discussion. Your presentation should cover five components: a summary of the application and its broader context, a summary of the method, a critique, proposed extensions, and discussion questions for the class.
Rubric (100 points total)
Summary of the application and its broader context (25 points)
- Mature framing of the problem gap within the domain, including a clear connection to the climate impact of filling that gap
- Mature framing of the methodological gap and an explanation of how it connects to the problem gap
- Clear communication that establishes common ground for peers, including introduction of key technical terms and concepts (e.g., what was unclear when you first encountered this work)
Summary of the method (25 points)
- Method is described accurately and in sufficient detail
- Method is contextualized within prior literature (e.g., how it differs from existing approaches and what is novel or significant about it)
Critique (25 points)
- In-depth analysis of aspects that were done poorly, considering: problem motivation, problem framing, methods development, and evaluation
- In-depth analysis of aspects that were done well using the same perspectives
Extensions (10 points)
- Proposal of one to two non-trivial extensions that demonstrate a holistic and nuanced understanding of the problem and method
- Extensions should go beyond simple modifications such as swapping in a new architecture and may include new problem owners, new methodological directions, new application contexts
Food for thought / questions for discussion (15 points)
- Three thoughtful discussion questions for the class
Final Project
The final project in this course will be to develop a proposal for a new AI-based research project for a climate application (we are very open to broad interpretations of both what "AI-based" and "climate application" entail). The proposal should concretely describe your idea and additionally include some initial validation that this proposed idea makes sense (e.g., targeted literature review; concrete discussion of data, proposed methods, and evaluation; and preliminary experiments). Think of this like a grant proposal to a funding agency, where you do not yet need to have completed the idea, but want to put forward a convincing case that the idea makes sense and should be "funded" to proceed. We would like to see proposals that are clearly motivated by the climate/biodiversity need, as well as clearly grounded in the existing AI and climate/biodiversity literature. We take the stance that novelty can take many forms, and are excited to see your proposed versions of "Application-driven innovation".
The components of the project will be:
- A preliminary presentation of the project at the end of the forked section of the course (to be presented in class April 13–22)
- A project proposal document due at the end of the course (May 11 by EOD)
You can work solo or in a group of up to 2. We will ask you to declare your groups by March 6.
Preliminary Presentation
The preliminary presentation will be an initial opportunity to share your project in progress with the class, and get feedback ahead of the final project proposal. The preliminary presentation will be graded on a Complete/Incomplete basis (full points if you do the presentation), and largely serves as an opportunity to receive feedback on the project direction, relevant related work, and ideas for methods and applications. It should be 15–20 minutes and include the following sections:
- Introduction
- Motivate the problem that your research project aims to solve
- Related Work
- Clearly outline the past work that relates to the proposed research and characterizes gaps that the method will hopefully fill
- Datasets
- Outline any potential sources of data that you are considering to include in the proposal, any pros and cons, or any existing data gaps
- Method and Experiment Design
- Describe your proposed research project methodology, including if and how you will build on prior work
- Describe any preliminary experiments/data analysis, or at minimum a plan for what experiments and analyses you will include in the final proposal (including what metrics you will measure)
- Broader Impacts
- Include a brief discussion of the pathway to impact for the project (what it would take to deploy, who the user/stakeholder community would be, any additional resources needed for translation)
Final Project Proposal
The final project proposal should be 6–8 pages long and include the following sections:
- Introduction
- Motivate the problem that your research project aims to solve
- Include at least one "hero" figure that captures the project goals at a high level
- Related Work
- Clearly outline the past work that relates to the proposed research and characterizes gaps that the method will hopefully fill
- Datasets
- Clearly define the data that you will work with for this project. If working from existing public datasets, justify why they are reasonably representative for the task or propose extensions/adaptations of those datasets. You can also propose the collection of new data here.
- If you will be proposing relabeling of additional data or collection of new data, we would like to see an estimate of the cost and time associated
- Method and Experiment Design
- Describe your research project methodology and design at least one figure that captures the method
- Describe if and how you will build on prior work
- Propose how you will test your method, what ablations you plan to do, and what metrics you will measure, as well as possibly including mockup figures of the trends you expect to see. Be concrete.
- Share the results of preliminary experiments aimed at validating the proposed approach. (We are open to broad interpretations of what "experiment" means depending on the context of your particular problem, data, and method.) Experiments may aim to, e.g., (a) show initial success on a smaller or "toy" version of the problem, or (b) identify the key risks or challenges in the proposed method, and justify that the proposed approach has the potential to overcome these key risks or challenges.
- Broader Impacts
- Finish by describing the potential impact the proposal could have on your application area, and describe what additional steps would need to be taken to translate the research to a user community
Final Project Rubric (100 points total)
1. Introduction & Motivation (20 pts)
- Problem is compelling, firmly rooted in a climate need. (10 pts)
- Benefits of an AI approach are quantified or evidenced. (5 pts)
- "Hero" figure is visually clear and immediately conveys end-to-end goals. (5 pts)
2. Related Work & Gap Analysis (15 pts)
- Survey demonstrates an understanding of the most closely related AI & climate/biodiversity literature to the proposed project. (5 pts)
- Gaps are clearly articulated (limitations of prior approaches, the need for this new problem formulation, etc.), positioning the proposal as a meaningful next step. (10 pts)
3. Dataset Plan & Justification (10 pts)
- Data sources (public or new) are precisely described, with representativeness, biases, and limitations analyzed. (10 pts)
- If the collection of new data and/or labels is part of the proposal: cost, logistics, time, ethics & permissions addressed.
4. Methodology & Experimental Design (35 pts)
- Method diagrams clearly communicate pipeline or model architecture. (6 pts)
- Builds logically on cited work with justified modifications/innovations. (5 pts)
- Evaluation plan lists metrics, baselines, and ≥1 ablation; mock-up plots/tables illustrate expected trends. (6 pts)
- Risks, limitations, and risk mitigation plan acknowledged. (5 pts)
- Preliminary experiments provide validation for the proposed approach, e.g., by showing initial success on a smaller or "toy" version of the problem, or by "derisking" major project risks. (13 pts)
5. Broader Impacts (10 pts)
- Climate, ecological, and societal benefits clearly articulated. (4 pts)
- Pathway to deployment (stakeholders, policy, outreach) outlined. (3 pts)
- Potential negative impacts or ethical concerns addressed. (3 pts)
6. Writing Quality & Presentation (10 pts)
- 6–8 pages, well-organized, concise writing, consistent citation style. Figures/tables are legible, captioned, and referenced in text. Follows all submission guidelines (file format, deadline, group size unless exception granted). (10 pts)