Big-Data Mechanisms and Energy-Policy Design
Authors: Ankit Pat, Kate Larson, Srinivasen Keshav
AAAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this paper we present an approach we call big-data mechanism design which combines a mechanism design framework with stakeholder surveys and data to allow policy-makers to gauge the costs and benefits of potential policy decisions. We illustrate the effectiveness of this approach in a concrete application domain: the peaksaver PLUS program in Ontario, Canada. and using a population of agents with utility functions extracted from our survey data, we ran a series of experiments to determine how different policies changed participation rates and what temperature changes we could support with which incentive policies. |
| Researcher Affiliation | Academia | Ankit Pat, Kate Larson and S. Keshav Cheriton School of Computer Science University of Waterloo Waterloo, ON, Canada {apat, klarson, keshav}@uwaterloo.ca |
| Pseudocode | No | No pseudocode or algorithm blocks are present in the paper. |
| Open Source Code | No | The paper does not provide concrete access to source code for the described methodology. |
| Open Datasets | No | To elicit agents preferences we conducted a survey to learn how agents react to the different non-cash incentives and to derive information about agents thermal discomfort and preferred baseline temperature. We used Crowd Flower1 for recruiting participants (with the geographical restriction that they must be in Canada or the United States), Survey Monkey2 to host the survey, and paid 1 USD to each participant. We had 990 responses and after a data cleaning process where we removed incomplete responses, duplicates based on IP filtering, responses that failed to follow directions, and those with very low variance across all questions, we were left with 425 responses. No explicit link or formal citation for the dataset itself. |
| Dataset Splits | No | using a population of agents with utility functions extracted from our survey data, we ran a series of experiments to determine how different policies changed participation rates and what temperature changes we could support with which incentive policies. The paper does not mention any train/validation/test splits for their data. |
| Hardware Specification | No | The paper does not specify any hardware details used for running experiments or computations. |
| Software Dependencies | No | The paper mentions 'Crowd Flower' and 'Survey Monkey' for data collection platforms but does not provide specific software dependencies with version numbers for replicating the experimental setup or analysis. |
| Experiment Setup | Yes | The Principal Let I be the set of (non-cash) incentives the principal may offer, and let θi {0, 1} be a variable where θi = 1 if the principal offers incentive i I and θi = 0 otherwise. We additionally assume that when the principal offers some incentive i I, it incurs a cost Ci. An incentive policy is specified by a vector P in = (θ1, . . . , θ|I|) indicating which incentives are offered by the principal. The cost associated with a particular incentive policy is cost(P in) = i I Ciθi. A policy, P, is an incentive policy coupled with a (homogeneous) temperature increase, ΔT, the principal wishes to implement. That is, P = P in, ΔT . We assume that there is some minimal level of agent participation the principal requires before it will implement a policy, and we denote this minimum threshold as λ . Given a policy, P, we let λ(P) be the number of agents willing to participate in the program given the policy. Given these conditions, we can specify the principal s optimization problem: maximize P λ(P)ΔT i Ciθi subject to λ(P) λ . (1) and For our model we leverage the significant body of research on determining thermal comfort for humans. Fanger proposed the Predicted Mean Vote (PMV) model to predict human comfort in different scenarios (Fanger 1970)... In particular, we assume humidity is equal to 50%, the mean radiant temperature is equal to the air temperature, the airspeed is zero, the metabolic rate is 1.2 met, and clothing level is 0.5 clo which represents typical summer indoor clothing (Hoyt et al. 2013). Using these assumptions, we simplify the PMV model using a linear regression model. Under this specific criteria, the regressed line fits PMV (ΔT) = 0.306|ΔT| 0.48 with mean square error 0.02%, where ΔT is the temperature change in Celsius. |