Lessons for Climate Change
Setting a Target to Increase or Reduce Carbon-Dioxide Concentration is Not the Same: Lessons for Climate Change Varun Dutt* and Cleotilde Gonzalez Dynamic Decision Making Laboratory Carnegie Mellon University 5000 Forbes Avenue, 208 Porter Hall Pittsburgh, PA 15213 *(412) 628-1379 (phone) / (412) 268-6938 (fax) [email protected] / [email protected] Abstract It has been observed that Kyoto protocol has defined CO2 emissions reduction goals that are less than what they should be. On the other hand, Intergovernmental Panel on Climate Change has set CO2 concentration stabilization goals that are definitely attainable in the near future but much above the current CO2 concentration than being below. In this study, we investigate reasons behind these real world observations in climate policies using an interactive simulation in a relevant climate change context (Dynamic Climate Change Simulator or DCCS). DCCS was used in a laboratory experiment to test participants’ control of CO2 concentration to a realistic goal over 200 simulated years. Participants confronted two different but symmetric initial values of CO2 concentration stock that provided different degrees of task difficulty: below the goal (minimum difficulty), and above the goal (maximum difficulty). Results show that on account of task difficulty participant performance remained poor under all conditions in the task when compared to the optimal performance. Also, participants starting above the goal had poorer performance than those starting below the goal. Reasons for these results were found in different from optimal participant emissions trajectories across different conditions. Real world policy implications of this research are discussed. Keywords: Dynamic decision making; climate change; stock and flows; stock-flow failure; goal; misperception of feedback; Introduction The Kyoto Protocol has set binding targets for 37 industrialized countries including the European community for reducing greenhouse gas (GHG) emissions. As per Kyoto, this will require an average of only a five per cent reduction in man-made carbon-dioxide (CO2) emissions against 1990 levels over a five-year period from 2008 to 2012 (United Nations, 1998). The reduction task requires the dominant world economies to reduce the CO2 emissions to only a little below the 1990 CO2 emission levels (rather than have a major reduction in their current CO2 emissions). But for the industrialized nations, even this small a reduction in emissions has currently become a major problem. As it has been reported recently, the Kyoto Protocol has become a failure and industrialized nations have not been able to achieve the intended reductions whatsoever (Bryner, 2007; Prins and Rayner, 2007; Victor, 2004).
1
On the other hand, the Intergovernmental Panel on Climate Change (IPCC) has set CO2 stabilization goals that are definitely attainable in the near future. But these goals are all above the current world CO2 concentration rather than being below the current world CO2 concentration (IPCC, 2001; 2007). Also, IPCC does not hint a single scenario where the CO2 concentration is stabilized at a goal that is below the current CO2 concentration. In this sense, there seems to be a clear consensus that achieving a higher than the present day concentration goal in the future is somewhat easier than achieving a goal that is lower than the current concentration. This is especially true for climate systems where the emissions are higher than natural CO2 absorptions due to oceans and biomass, where there are restrictions on emission reductions and where the emissions constitute the only factor in direct control of human beings (IPCC, 2001; Sterman and Booth Sweeney, 2002, 2007). For our climate, the total manmade CO2 emissions in the 2000-2010 decade have been estimated to average 8 GtC/year (i.e. Giga or Billion tons of Carbon/year) and are increasing every year (IPCC, 2001; 2007). While the CO2 absorptions by natural processes like the oceans and biomass (considering them to be proportional to the CO2 concentrations) are found close to 1 GtC/year (IPCC, 2001, 2007). This creates an imbalance where the current emissions which add CO2 to the atmosphere are far in excess and greater than the CO2 absorptions. Due to this effect, the CO2 concentration is currently on a rapid rise. Also, there are many social, economic and political reasons which motivate the dominant world economies unwillingness to cut down the CO2 emissions (IPCC, 2007; Sterman and Booth Sweeney, 2002, 2007). As a result, there are a number of narrow bounds that can be set in current IPCC’s Special Report Emission scenarios (SRES; Jain, Kheshgi, and Wuebbles, 1994) where the scenarios depict future CO2 emission storylines. From the SRES one can estimate the heavy restrictions put on reduction of two dominant manmade emissions, fossil fuels and deforestation in the current SRES (Dutt & Gonzalez, 2008; IPCC 2001). Also, it is a well known fact that in the climate system, humans can only exert direct control on man-made CO2 emissions and cannot directly manipulate the absorptions of CO2 which are due to natural causes (IPCC, 2007; Sterman and Booth Sweeney, 2002, 2007). In this paper we aim at demonstrating that the IPCC view of setting goals that are above our current CO2 concentration level leads to an easier and more successful control of the CO2 concentration, in contrast to the Kyoto proposal of setting goals below the current emission level. We will demonstrate this proposal by using an abstraction of the complex climate system in a Dynamic Climate Change Simulator (DCCS) (Dutt & Gonzalez, 2008). To understand the complex dynamics behind this proposition we can present an analogy of a speed boat in a flowing river where the speed of the boat is CO2 emissions and the speed of the river is the CO2 absorptions. When the current CO2 concentration is above an attainable goal and CO2 emissions are greater than CO2 absorptions (which they are currently as per Dutt and Gonzalez, 2008; Sterman and Booth Sweeney, 2002, 2007), it is just like travelling in the boat against the current flow i.e. travelling upstream. The boat needs to exceed the velocity of the river to be able to make progress towards an upstream goal or destination. Thus, one would need to reduce the CO2 emissions below the CO2 absorptions to be able to make a movement towards an attainable goal set below the current CO2 concentration. In contrast, to move in the river towards a goal or destination which is downstream i.e. in the direction of current flow, one would not need to exceed the velocity of the river, as in this case the river would help the boat to reach the goal even when the engines are turned off. In fact the boat might only need to slow down to stay at the downstream goal. Thus, for the climate problem, starting from a CO2 concentration below the
2
goal to reach a higher attainable goal, one needs to maintain the CO2 emissions above the CO2 absorptions (as has been in the status quo) and make them only equal the CO2 absorptions when one is about to attain the goal to stay at the goal. Given that delays and relationships involved between the CO2 emissions and their effect on the CO2 concentration, and the delays and relationships involved between the concentration and CO2 absorptions, when the CO2 emissions are higher than the CO2 absorptions, it is very difficult to reduce the CO2 concentration in the short-term near future; particularly given other social, economic, and political reasons (Dutt and Gonzalez, 2008; IPCC, 2007). Thus, it would require less effort to increase the CO2 concentration to a goal set higher than the current concentration than it would be to decrease the CO2 concentration to a goal set lower than the concentration. Specifically, using the Dynamic Climate Change Simulator (DCCS) tool which is a management flight simulator that we have used extensively in the past to test human perceptions on questions related to climate change (Dutt & Gonzalez, 2008), in this paper we will determine whether people’s control of CO2 concentration to a goal is poorer when they start above an attainable goal and reduce the CO2 concentration to a goal than when they start below the goal and increase the CO2 concentration to a goal. We use a lab based approach for the purpose of our investigation using human participants in an experiment. In the next section, we will briefly describe DCCS (Dutt & Gonzalez, 2008). Then, we will present a lab based experiment to test our hypotheses. Lastly, we close with a discussion of our results and their application to real world climate problems and policy making. Dynamic climate change simulator (DCCS) The model which was used to build DCCS was explained in Dutt and Gonzalez (2008). Thus, here we provide only a brief summary of this model. Figure 1 provides the system dynamics representation of our climate model. The CO2 in Atmos represents concentration of CO2 in the atmosphere (i.e., stock). The CO2 concentration increases indirectly by human decisions on anthropogenic CO2 emissions (i.e., inflow) called User Action CO2 emissions (thus here User Action CO2 emissions are made up of 2 kinds of emissions: fossil fuel and deforestation types). The CO2 emissions into the Atmosphere are only affected by User Action CO2 emissions which in turn increase the stock of CO2 in Atmos. The CO2 absorptions (i.e., outflow) cause a decrease in the concentrations of the CO2 in Atmos stock due to absorptions by terrestrial and ocean ecosystems. As long as the CO2 emissions into the Atmosphere or User Action CO2 emissions exceed absorption rates, i.e. CO2 absorptions, the CO2 in Atmos continues to increase. Only when the emissions equal the absorption rates will the CO2 in Atmos be stabilized. The arrow from the CO2 in Atmos to the CO2 absorptions illustrates that the outflow at all times depends on the concentration of CO2 in Atmos. For the equation, CO2 absorptions are directly proportional to the concentration of CO2.
3
Fig. 1. Climate Model The Rate of CO2 Transfer is a constant multiplier into CO2 in Atmos that gives rise to CO2 absorptions after the Pre-industrial CO2 (1970 baseline CO2 concentration) has been subtracted from the CO2 in Atmos. The use of a baseline concentration and year enables us to determine the change in value of CO2 absorptions in comparison to a common starting point or datum. Equations used in the model can be found in supplementary material of this paper. As per results reported in Dutt and Gonzalez (2008) we found that participant performance was the worst when participants confronted the condition of climate dynamic where Rate of CO2 Transfer had values of 0.012 (or 1.2%) of the CO2 concentration per year. In this experiment we use the same climate dynamics condition and fix the rate of CO2 transfer parameter to get a benchmark on participant performance due to other manipulated factors. The DCCS interface (see Figure 2) represents a single stock or accumulation of CO2 in the form of an orange-color liquid in a tank (number 1). Anthropogenic deforestation and fossil fuel CO2 emissions, are represented by a pipe connected to the tank (number 2), that increase the level of CO2 stock; and CO2 absorptions, also represented as a pipe on the right of the tank (number 3), which decreases the level of CO2 stock.
Fig. 2. Dynamic Climate Change Simulator task
4
The participant’s goal in DCCS was to maintain the CO2 concentration within an attainable and acceptable range around the goal value of 938 GtC shown with a green horizontal line with a label Goal. The participant decided on emissions of two different types: deforestation and burning of fossil fuels (number 4). The participant set the fossil fuel and deforestation emissions in the respective text boxes labeled Fossil Fuel Emissions (GtC/year) and Deforestation Emissions (GtC/year), and click the Make Emission Decision button. To avoid extreme exploration of participants’ emission decisions, we restricted the fossil fuel and deforestation emissions values to between From and To ranges calculated after each emission decision time step executed by the participant (number 5; for details on calculations of the ranges see supplementary material). When a participant makes a decision DCCS moves autonomously by a number of time steps to a future year. As per the results reported in Dutt and Gonzalez (2008) we found that participant performance was the worst when participants confronted the situation where the system moved 4 time steps to a future year. To get a lower bound on participant performance in the current experiment, we assumed the same emission decision frequency and fixed the frequency of emissions decisions parameter at the value of every 4 years in DCCS. For other details on DCCS please see Dutt and Gonzalez (2008). In the next section, we describe a laboratory experiment conducted using DCCS. Experiment Using DCCS, we manipulated the initial or starting value of CO2 concentration in the system to two different levels, i.e. above the goal and below the goal. The purpose of the experiment is to study human decision making and control for determining implications towards real world climate policies for different degrees of difficulty that is created by different initial CO2 concentrations relative to an attainable goal. To accomplish our intended purpose, as explained earlier, we took a scenario where we fixed the rate of transfer at 1.2% per year of CO2 concentration. Also, we fixed emission decision frequency to 4 years in DCCS. We expected participants to show better control of CO2 to an attainable goal when they start below the goal, than when participants start above the goal. This effect would be primarily attributed to a participant’s poor and deteriorating control performance due to task difficulty of the climate problem when people start above an attainable goal than when they start below the goal. Starting above the goal than starting below the goal is more challenging as for climate systems, current emissions much higher than absorptions, there is a positive relationship between the CO2 emissions and CO2 concentration, and there are tight restrictions on current CO2 emission reductions in the world. Methods Experimental Design Participants were randomly assigned to one of two initial CO2 concentration conditions: above the goal (Above) and below the goal (Below). An attainable goal under all initial CO2 concentration conditions was to maintain the level of CO2 within +/- 15 GtC from the 938 GtC goal (i.e. 923 GtC to 953 GtC and taken from the 450 ppmv definition of goal by IPCC, 2001). Participants assigned to Above and Below conditions had their initial CO2 concentration set at 1107 GtC and 769 GtC respectively. The values 769 GtC and 1107 GtC are equidistant from the 938 GtC goal while the value of 769 GtC constitutes the actual world value of CO2 concentration 5
in year 20001. The participants started DCCS in the year 2000. We ran the climate change scenario, just described, for 200 years and analyzed the two experiment conditions using 50 decision points i.e. using every 4th year from the start year for data analysis (this filtering of data points helped us avoid repetition of same emissions decisions 4 times). The From and To ranges of fossil fuel emissions were set at -14% to +22% of the value of current fossil fuel emissions. Also, for deforestation emissions, the From and To ranges were set as -51% to +55% of the value of current deforestation emissions (details on how these values were obtained can be read from the supplementary material of this paper). These ranges serve the purpose of providing realistic bounds on possible increases and decreases in fossil fuel and deforestation emissions by participants. We used the CO2 concentration, total emissions, absorptions and decision point to first reach the goal as dependent variables to discuss participant decision making and performance in this experiment. Participants Sixteen graduate and undergraduate students from diverse fields of study participated in this experiment. 6 were females, and 10 were males. Ages ranged from 18 years to 24 years with an average age of 22 years and standard deviation of 1.35 years. 68% of the participants selfreported to have degrees in STEM. Six participants were randomly assigned to the Above condition and ten to the Below condition. All participants received a base pay of $5. The participants could earn an additional bonus out of a maximum of $3, based on their performance in the DCCS task. The calculation of performance bonus was tied to the cost incurred by a linear relationship. Each time a participant could not keep the CO2 stock between +/- 15 GtC above or below the goal of 938 GtC in DCCS (i.e. the CO2 stock level did not fall between 923 GtC and 953 GtC), the participant incurred a cost which was $100 million times the discrepancy or difference between the upper or lower bound of the goal and the CO2 concentration stock. Cost incurred was $0 if the participant was able to control the CO2 concentration stock to stabilize it between 923 GtC and 953 GtC. Participants got $3 in performance bonus if they could keep the total costs in the task to be less than or equal to $15 billion. Participants got $0 in performance bonus (i.e. only got the base pay) if they incurred a total cost of $400 billion or more. For total costs between $400 billion and $15 billion, the cost was linearly translated to give a bonus which fell between $0 and $3. Procedure Participants were randomly assigned to one of the two conditions (Above or Below), and they were given instructions on the computer before starting the DCCS task. Text of the instruction for the Below condition is detailed in supplementary material of this paper with pointers to changes in text for other two conditions. Soon after participants read the instructions, they were shown a 50 year status quo scenario in DCCS. This scenario demonstrated what the state of affairs would be if no action had been taken at all in the simulation. The scenario started in the year 2000 and ended in year 2050. In this scenario, the year 2000 emissions for deforestation were maintained at 1.3 GtC per year and the emissions for fossil fuels at 6.88 GtC per year for the next 50 years. For people put on the Above and Below conditions the initial concentration was kept at 1107 GtC and 769 GtC in the 50 year status quo scenario. If we start with year 2000 1
Recorded at the Mauna Loa observatory, Hawaii: http://www.esrl.noaa.gov/gmd/ccgg/trends/
6
CO2 concentration of 769 GtC i.e. Below condition, the concentration of CO2 in the atmospheric tank increased to more than 1000 GtC by year 2050, which is more than a 5% increase from the year 2000 value (this can be derived from the climate model detailed earlier). Participants were told of the severe consequences that this increase in concentration might hold in terms of economic and life losses under all the conditions of the experiment. After the scenario finished, participants were encouraged to ask questions on the task. Participants were given full information on the DCCS task and were reminded that their task was to control CO2 concentration to within the +/- 15 GtC range around the goal over the entire course of the DCCS task. They were then asked to play the DCCS task for 50 decision points over a course of 200 years either starting in Above or Below the goal conditions. Participants sat in front of a computer screen and were presented with the DCCS displayed in the center of the screen. We made use of Windows XP-based desktop computer terminals with plasma screens, which had a resolution of 1024 x 1080; the desktop computers ran on an Intel® Pentium 4 processor base. Optimal solution to climate change This section details optimal solutions for the two climate change conditions Above and Below using the dynamics we described for rate of CO2 transfer, frequency of emissions decision and other parameters in the preceding sections. These optimal solutions were developed using two different starting point of CO2 concentrations in the climate model detailed earlier. The model captured the basic elements of a climate system i.e. emissions much higher than absorptions, a positive relationship between the CO2 emissions and CO2 concentration, and tight restrictions on current CO2 emission reductions. It must be mentioned that although the values of CO2 concentration has been chosen in Above and Below conditions such that the CO2 concentrations are symmetrically around a realistic 938 GtC (~450 ppmv) goal (IPCC, 2001) these values actually create different task difficulty as described next specific to each condition of the experiment. Below Condition In the Below condition, as shown in Figure 3, the starting CO2 concentration is below an attainable goal of 938 GtC. The optimal solution (those that control the system in the minimum number of decision points) is obtained by increasing the fossil fuel and deforestation emissions at a maximum rate i.e. by maintaining their values at the To value ranges till the concentration approaches the goal. Once the concentration reaches goal the fossil fuel and deforestation emissions are reduced at a maximum rate i.e. by maintaining their values at the From range values till they equalize the CO2 absorptions (to stabilize CO2 concentration at the goal). It takes about 5 decision points to reach the goal range between 923 GtC and 953 GtC in Below condition. This will correspond to 20 (=5*4) years in DCCS.
7
Fig. 3. Optimal Total CO2 Emissions, Absorptions and CO2 concentration in the Below condition Above Condition In the Above condition, shown in Figure 4, the starting CO2 concentration is above an attainable goal of 938 GtC. The optimal solution is obtained by firstly decreasing the fossil fuel and deforestation emissions at a maximum rate i.e. by maintaining their values at the From range value till the concentration approaches the goal (emissions are below the absorptions and hence the concentration diminishes). Once the concentration reaches goal, the fossil fuel and deforestation emissions are increased at a maximum rate i.e. by maintaining their values at the To values of their respective ranges till they equalize the CO2 absorptions (this process will stabilize CO2 concentration at the goal). It takes about 18 decision points to reach the goal range between 923 GtC and 953 GtC in the Above condition. This will correspond to 72 (=18*4) years in DCCS.
8
Fig. 4. Optimal Total CO2 Emissions, Absorptions and CO2 concentration in the Above condition Thus, it becomes clear that as per the computed optimal strategies, the Below condition takes less number of decision points and hence it is easier than the Above condition in task difficulty where the Above condition takes more number of decision points to reach the goal. The next section details results of participant CO2 emissions decisions compared to optimal strategies reported in this section. Results Figure 5 and Figure 6 show the fits of optimal emissions derived in the last section to average human behavior shown for the resulting CO2 concentration, total emissions and absorptions obtained from our experiment in Below and Above conditions respectively. It can be seen from Figure 5 and Figure 6 that the participants did well to control the CO2 concentration to goal when they started Below than when they started Above (higher R2 and lower RMSD for Below condition than Above condition). Also, after comparing the average total CO2 emissions and average CO2 absorptions in Figure 5, we see that participants come close to the optimal
9
trajectories when starting below the goal. Whereas, as reported in Figure 6, the participant trajectories for average total CO2 emissions and average CO2 absorptions are similar in trend or shape to the optimal trajectories (R2 ~ 0.75) but they are highly deviated from the optimal trajectories (very high RMSD values). Also the trend in Above condition is not as good as the trend for average total CO2 emissions and average CO2 absorptions in the Below condition. These results show that participants average total CO2 emissions decisions were better when they started below the goal than when they started above. Also, similar comparisons can be made for average CO2 concentration, and average CO2 absorptions where these measures had a better fit to optimal for the Below condition than the Above condition.
10
Fig. 5. CO2 concentration, total emissions and absorptions from optimal solution and humans in the Below condition across 50 decision points (human curves are averaged over all participants in Below condition for each decision point). Error bars show 90% confidence interval around the point estimate. The R2 and RMSD are reported in each figure for fits of CO2 concentration, total emissions and absorptions of optimal solution to humans.
11
12
Fig. 6. CO2 concentration, total emissions and absorptions from optimal solution and humans in the Above condition across 50 decision points (human curves are averaged over all participants in Above condition for each decision point). Error bars show 90% confidence interval around the point estimate. The R2 and RMSD are reported in each figure for fits of CO2 concentration, total emissions and absorptions of optimal solution to humans. Time to Reach the Goal Here we analyze the number of decision points on average that participants took to reach the goal for the first time. The analysis of decision point to first reach the goal indicated that participants in the Below condition reached the goal faster (Mean = 16.46 decision points; SE = 3.47 decision points) than participants in the Above condition (Mean = 27.75 decision points; SE = 11.98 decision points). Figure 7 reports the comparison of Above and Below conditions for humans and optimal solution. Humans in both Above and Below conditions are far away from the optimal solutions in the number of decision points to reach the goal. The humans take less number of decision points to get CO2 concentration to reach the goal range for Below as compared to Above condition and likewise is the case for the optimal solutions for the two conditions.
Fig. 7. Average Decision Point to reach the goal under Above and Below conditions for humans and optimal solutions. The results show that the Humans were away from the optimal by about 11 Decision Points across the Above and Below Conditions (RMSD ~ 11). Error bars show 90% confidence interval around the point estimate. Discussion and conclusions Many of the complex dynamic effects found in the real world (like CO2 dynamics) can be better understood with simple tasks (Cronin et al., 2009) and here we present a demonstration of such a process in DCSS. The results reported in this paper clearly support an argument in favor of our initial expectation of poorer control performance in a system where participants decrease their
13
emissions in order to reduce concentration towards an attainable goal, than in a system where they increase their emissions in order to increase concentration towards the same goal. Our results using different dependent measures indicate that participants starting above the goal exhibited poor control performance than those starting below the goal (lower R2 and higher RMSD values for humans when compared with optimal for Above than Below condition). Our results also reveal that the participant control performance as seen in their CO2 concentration’s control to goal in DCCS generally remained poor although the control was better and similar to the optimal solution in the Below condition than the Above condition. Further, participants made poor selection of average total emissions decisions in the Above than in the Below condition when both conditions were compared with the optimal solution. The deterioration of control performance for starting above the goal than below is due to one simple reason: task difficulty on account of basic climatic factors where emissions are greater than absorptions, emissions are restricted to upper and lower bounds and emissions are the only decision making elements directly controlled by the decision makers. Research Implications This research provides interesting cognitive explanations for real world climate policy making that involve reducing emissions as per the Kyoto Protocol and those of IPCC’s higher than current concentration goal setting practices. Because people find it harder to understand, perceive and control the system dynamics for our climate when they are above the goal than when they are below, on account of task’s difficulty and their cognitive limitations, they prefer to set goals that are higher relative to where they stand today. Although, there could be many economic, social and political reasons yet one simple cognitive reason explains why Kyoto has been a failure and why IPCC sets goals that are higher relative to where we stand today – people’s tendency to easily perceive and prefer challenges that are less difficult than those that are more difficult to attain. The main policy implication from this research is that there is currently some bad news in terms of future of policy making for our climate from a purely cognitive view point. People want to anchor and adjust from their current positions relative to where they stand but prefer to do it in a direction that provides them a path to lessen the problem’s difficulty i.e. where their deficient mental representations are more easily satisficed (Simon, 1990; Tversky, and Kahneman, 1974). This anchor and adjustment and satisficing behavior for a dynamic system like the climate from our current research results means that people will have a constant tendency to look upwards and set higher goals as the world attains even higher levels for the rising atmospheric CO2 concentrations in the future. This behavior on part of humans will prove extremely detrimental for our climate. Today we are crossing the 800 GtC CO2 concentration mark and still going up (as per the Mauna Loa 2009 estimates2) and because we are cognitively able to think upwards easily than downwards, we are sadly thinking of making higher stabilization plans which are in excess of 1000 GtC today. We plan to continue our initial research investigations reported in this paper. In particular, it is important to understand how the dynamic complexity posed by the climate dynamics (which was assumed fixed in this research) would interact with the task difficulty of different initial or starting CO2 concentrations. For example, it would be important to see how rapid dynamics, where the climate removes CO2 more rapidly from the atmosphere (i.e. at a higher Rate of CO2 transfer), helps people controlling the climate system better even when people start above the 2
http://www.esrl.noaa.gov/gmd/ccgg/trends/
14
goal. As for another follow up experiment, it would be good to determine if our current hypothesis on task difficulty will become true for situations where people start at the same initial value of CO2 concentration in all conditions. But now have to control CO2 concentration to two symmetric and attainable goals from the common starting point, one of which is above and the other below the common starting point. We see that greater understanding of human incapacities and cognitive limitations using tools like DCCS is the right step for research in this endeavor. Acknowledgments This research was partially supported by the National Science Foundation (Human and Social Dynamics: Decision, Risk, and Uncertainty, Award number: 0624228) award to Cleotilde Gonzalez.
15
REFERENCES Bryner G. 2007. Lessons from the Kyoto Protocol for the Design of Future Global Climate Accords. In Proceedings, the annual meeting of the International Studies Association 48th Annual Convention. Chicago, IL. Cronin M, Gonzalez C, Sterman JD. 2009. Why don't well-educated adults understand accumulation? A challenge to researchers, educators and citizens. Journal of Organizational Behavior and Human Decision Processes 108(1): 116-130. Dutt V, Gonzalez C. 2008. Human Perceptions of Climate Change. In Proceedings, 26th International Conference of the System Dynamics Society. Athens, Greece. [CD-ROM] IPCC. 2001. Climate Change 2001: The Scientific Basis. Intergovernmental Panel on Climate Change 83. IPCC. 2007. Climate Change 2007: Synthesis Report. Contribution of Working Groups I, II and III to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change [Core Writing Team, Pachauri, R.K and Reisinger, A. (eds.)]. IPCC. Geneva, Switzerland, pp. 104. Jain AK, Kheshgi HS, and Wuebbles DJ. 1994. Integrated Science Model for Assessment of Climate Change. In Proceedings, Air and Waste Management Association’s 87th Annual Meeting. Cincinnati, Ohio. Prins G, Rayner S. 2007. Time to ditch Kyoto. Nature 449: 973-975. Simon HA. 1990. Invariants of Human Behavior. Annual Review of Psychology 41: 1-19. Sterman JD, Booth Sweeney L. 2002. Cloudy skies: Assessing public understanding of global warming. System Dynamics Review 18(2): 207-240. Sterman JD, Booth Sweeney L. 2007. Understanding public complacency about climate change: Adults' mental models of climate change violate conservation of matter. Climatic Change 80(3-4): 213-238. Tversky A, Kahneman D. 1974. Judgment under uncertainty: Heuristics and biases. Science 185(4157): 1124-1131. United Nations. 1998. Framework Convention on Climate Change. Kyoto Protocol. Climate Change Secretariat (UNFCCC). HTTP: http://unfccc.int/kyoto_protocol/items/2830.php (Retrieved on 17th March 2009). [Online]. Victor DG. 2004. The Collapse of the Kyoto Protocol and the Struggle to Slow Global Warming (Council on Foreign Relations Book). Princeton, MA: Princeton University Press.
16
1
On the other hand, the Intergovernmental Panel on Climate Change (IPCC) has set CO2 stabilization goals that are definitely attainable in the near future. But these goals are all above the current world CO2 concentration rather than being below the current world CO2 concentration (IPCC, 2001; 2007). Also, IPCC does not hint a single scenario where the CO2 concentration is stabilized at a goal that is below the current CO2 concentration. In this sense, there seems to be a clear consensus that achieving a higher than the present day concentration goal in the future is somewhat easier than achieving a goal that is lower than the current concentration. This is especially true for climate systems where the emissions are higher than natural CO2 absorptions due to oceans and biomass, where there are restrictions on emission reductions and where the emissions constitute the only factor in direct control of human beings (IPCC, 2001; Sterman and Booth Sweeney, 2002, 2007). For our climate, the total manmade CO2 emissions in the 2000-2010 decade have been estimated to average 8 GtC/year (i.e. Giga or Billion tons of Carbon/year) and are increasing every year (IPCC, 2001; 2007). While the CO2 absorptions by natural processes like the oceans and biomass (considering them to be proportional to the CO2 concentrations) are found close to 1 GtC/year (IPCC, 2001, 2007). This creates an imbalance where the current emissions which add CO2 to the atmosphere are far in excess and greater than the CO2 absorptions. Due to this effect, the CO2 concentration is currently on a rapid rise. Also, there are many social, economic and political reasons which motivate the dominant world economies unwillingness to cut down the CO2 emissions (IPCC, 2007; Sterman and Booth Sweeney, 2002, 2007). As a result, there are a number of narrow bounds that can be set in current IPCC’s Special Report Emission scenarios (SRES; Jain, Kheshgi, and Wuebbles, 1994) where the scenarios depict future CO2 emission storylines. From the SRES one can estimate the heavy restrictions put on reduction of two dominant manmade emissions, fossil fuels and deforestation in the current SRES (Dutt & Gonzalez, 2008; IPCC 2001). Also, it is a well known fact that in the climate system, humans can only exert direct control on man-made CO2 emissions and cannot directly manipulate the absorptions of CO2 which are due to natural causes (IPCC, 2007; Sterman and Booth Sweeney, 2002, 2007). In this paper we aim at demonstrating that the IPCC view of setting goals that are above our current CO2 concentration level leads to an easier and more successful control of the CO2 concentration, in contrast to the Kyoto proposal of setting goals below the current emission level. We will demonstrate this proposal by using an abstraction of the complex climate system in a Dynamic Climate Change Simulator (DCCS) (Dutt & Gonzalez, 2008). To understand the complex dynamics behind this proposition we can present an analogy of a speed boat in a flowing river where the speed of the boat is CO2 emissions and the speed of the river is the CO2 absorptions. When the current CO2 concentration is above an attainable goal and CO2 emissions are greater than CO2 absorptions (which they are currently as per Dutt and Gonzalez, 2008; Sterman and Booth Sweeney, 2002, 2007), it is just like travelling in the boat against the current flow i.e. travelling upstream. The boat needs to exceed the velocity of the river to be able to make progress towards an upstream goal or destination. Thus, one would need to reduce the CO2 emissions below the CO2 absorptions to be able to make a movement towards an attainable goal set below the current CO2 concentration. In contrast, to move in the river towards a goal or destination which is downstream i.e. in the direction of current flow, one would not need to exceed the velocity of the river, as in this case the river would help the boat to reach the goal even when the engines are turned off. In fact the boat might only need to slow down to stay at the downstream goal. Thus, for the climate problem, starting from a CO2 concentration below the
2
goal to reach a higher attainable goal, one needs to maintain the CO2 emissions above the CO2 absorptions (as has been in the status quo) and make them only equal the CO2 absorptions when one is about to attain the goal to stay at the goal. Given that delays and relationships involved between the CO2 emissions and their effect on the CO2 concentration, and the delays and relationships involved between the concentration and CO2 absorptions, when the CO2 emissions are higher than the CO2 absorptions, it is very difficult to reduce the CO2 concentration in the short-term near future; particularly given other social, economic, and political reasons (Dutt and Gonzalez, 2008; IPCC, 2007). Thus, it would require less effort to increase the CO2 concentration to a goal set higher than the current concentration than it would be to decrease the CO2 concentration to a goal set lower than the concentration. Specifically, using the Dynamic Climate Change Simulator (DCCS) tool which is a management flight simulator that we have used extensively in the past to test human perceptions on questions related to climate change (Dutt & Gonzalez, 2008), in this paper we will determine whether people’s control of CO2 concentration to a goal is poorer when they start above an attainable goal and reduce the CO2 concentration to a goal than when they start below the goal and increase the CO2 concentration to a goal. We use a lab based approach for the purpose of our investigation using human participants in an experiment. In the next section, we will briefly describe DCCS (Dutt & Gonzalez, 2008). Then, we will present a lab based experiment to test our hypotheses. Lastly, we close with a discussion of our results and their application to real world climate problems and policy making. Dynamic climate change simulator (DCCS) The model which was used to build DCCS was explained in Dutt and Gonzalez (2008). Thus, here we provide only a brief summary of this model. Figure 1 provides the system dynamics representation of our climate model. The CO2 in Atmos represents concentration of CO2 in the atmosphere (i.e., stock). The CO2 concentration increases indirectly by human decisions on anthropogenic CO2 emissions (i.e., inflow) called User Action CO2 emissions (thus here User Action CO2 emissions are made up of 2 kinds of emissions: fossil fuel and deforestation types). The CO2 emissions into the Atmosphere are only affected by User Action CO2 emissions which in turn increase the stock of CO2 in Atmos. The CO2 absorptions (i.e., outflow) cause a decrease in the concentrations of the CO2 in Atmos stock due to absorptions by terrestrial and ocean ecosystems. As long as the CO2 emissions into the Atmosphere or User Action CO2 emissions exceed absorption rates, i.e. CO2 absorptions, the CO2 in Atmos continues to increase. Only when the emissions equal the absorption rates will the CO2 in Atmos be stabilized. The arrow from the CO2 in Atmos to the CO2 absorptions illustrates that the outflow at all times depends on the concentration of CO2 in Atmos. For the equation, CO2 absorptions are directly proportional to the concentration of CO2.
3
Fig. 1. Climate Model The Rate of CO2 Transfer is a constant multiplier into CO2 in Atmos that gives rise to CO2 absorptions after the Pre-industrial CO2 (1970 baseline CO2 concentration) has been subtracted from the CO2 in Atmos. The use of a baseline concentration and year enables us to determine the change in value of CO2 absorptions in comparison to a common starting point or datum. Equations used in the model can be found in supplementary material of this paper. As per results reported in Dutt and Gonzalez (2008) we found that participant performance was the worst when participants confronted the condition of climate dynamic where Rate of CO2 Transfer had values of 0.012 (or 1.2%) of the CO2 concentration per year. In this experiment we use the same climate dynamics condition and fix the rate of CO2 transfer parameter to get a benchmark on participant performance due to other manipulated factors. The DCCS interface (see Figure 2) represents a single stock or accumulation of CO2 in the form of an orange-color liquid in a tank (number 1). Anthropogenic deforestation and fossil fuel CO2 emissions, are represented by a pipe connected to the tank (number 2), that increase the level of CO2 stock; and CO2 absorptions, also represented as a pipe on the right of the tank (number 3), which decreases the level of CO2 stock.
Fig. 2. Dynamic Climate Change Simulator task
4
The participant’s goal in DCCS was to maintain the CO2 concentration within an attainable and acceptable range around the goal value of 938 GtC shown with a green horizontal line with a label Goal. The participant decided on emissions of two different types: deforestation and burning of fossil fuels (number 4). The participant set the fossil fuel and deforestation emissions in the respective text boxes labeled Fossil Fuel Emissions (GtC/year) and Deforestation Emissions (GtC/year), and click the Make Emission Decision button. To avoid extreme exploration of participants’ emission decisions, we restricted the fossil fuel and deforestation emissions values to between From and To ranges calculated after each emission decision time step executed by the participant (number 5; for details on calculations of the ranges see supplementary material). When a participant makes a decision DCCS moves autonomously by a number of time steps to a future year. As per the results reported in Dutt and Gonzalez (2008) we found that participant performance was the worst when participants confronted the situation where the system moved 4 time steps to a future year. To get a lower bound on participant performance in the current experiment, we assumed the same emission decision frequency and fixed the frequency of emissions decisions parameter at the value of every 4 years in DCCS. For other details on DCCS please see Dutt and Gonzalez (2008). In the next section, we describe a laboratory experiment conducted using DCCS. Experiment Using DCCS, we manipulated the initial or starting value of CO2 concentration in the system to two different levels, i.e. above the goal and below the goal. The purpose of the experiment is to study human decision making and control for determining implications towards real world climate policies for different degrees of difficulty that is created by different initial CO2 concentrations relative to an attainable goal. To accomplish our intended purpose, as explained earlier, we took a scenario where we fixed the rate of transfer at 1.2% per year of CO2 concentration. Also, we fixed emission decision frequency to 4 years in DCCS. We expected participants to show better control of CO2 to an attainable goal when they start below the goal, than when participants start above the goal. This effect would be primarily attributed to a participant’s poor and deteriorating control performance due to task difficulty of the climate problem when people start above an attainable goal than when they start below the goal. Starting above the goal than starting below the goal is more challenging as for climate systems, current emissions much higher than absorptions, there is a positive relationship between the CO2 emissions and CO2 concentration, and there are tight restrictions on current CO2 emission reductions in the world. Methods Experimental Design Participants were randomly assigned to one of two initial CO2 concentration conditions: above the goal (Above) and below the goal (Below). An attainable goal under all initial CO2 concentration conditions was to maintain the level of CO2 within +/- 15 GtC from the 938 GtC goal (i.e. 923 GtC to 953 GtC and taken from the 450 ppmv definition of goal by IPCC, 2001). Participants assigned to Above and Below conditions had their initial CO2 concentration set at 1107 GtC and 769 GtC respectively. The values 769 GtC and 1107 GtC are equidistant from the 938 GtC goal while the value of 769 GtC constitutes the actual world value of CO2 concentration 5
in year 20001. The participants started DCCS in the year 2000. We ran the climate change scenario, just described, for 200 years and analyzed the two experiment conditions using 50 decision points i.e. using every 4th year from the start year for data analysis (this filtering of data points helped us avoid repetition of same emissions decisions 4 times). The From and To ranges of fossil fuel emissions were set at -14% to +22% of the value of current fossil fuel emissions. Also, for deforestation emissions, the From and To ranges were set as -51% to +55% of the value of current deforestation emissions (details on how these values were obtained can be read from the supplementary material of this paper). These ranges serve the purpose of providing realistic bounds on possible increases and decreases in fossil fuel and deforestation emissions by participants. We used the CO2 concentration, total emissions, absorptions and decision point to first reach the goal as dependent variables to discuss participant decision making and performance in this experiment. Participants Sixteen graduate and undergraduate students from diverse fields of study participated in this experiment. 6 were females, and 10 were males. Ages ranged from 18 years to 24 years with an average age of 22 years and standard deviation of 1.35 years. 68% of the participants selfreported to have degrees in STEM. Six participants were randomly assigned to the Above condition and ten to the Below condition. All participants received a base pay of $5. The participants could earn an additional bonus out of a maximum of $3, based on their performance in the DCCS task. The calculation of performance bonus was tied to the cost incurred by a linear relationship. Each time a participant could not keep the CO2 stock between +/- 15 GtC above or below the goal of 938 GtC in DCCS (i.e. the CO2 stock level did not fall between 923 GtC and 953 GtC), the participant incurred a cost which was $100 million times the discrepancy or difference between the upper or lower bound of the goal and the CO2 concentration stock. Cost incurred was $0 if the participant was able to control the CO2 concentration stock to stabilize it between 923 GtC and 953 GtC. Participants got $3 in performance bonus if they could keep the total costs in the task to be less than or equal to $15 billion. Participants got $0 in performance bonus (i.e. only got the base pay) if they incurred a total cost of $400 billion or more. For total costs between $400 billion and $15 billion, the cost was linearly translated to give a bonus which fell between $0 and $3. Procedure Participants were randomly assigned to one of the two conditions (Above or Below), and they were given instructions on the computer before starting the DCCS task. Text of the instruction for the Below condition is detailed in supplementary material of this paper with pointers to changes in text for other two conditions. Soon after participants read the instructions, they were shown a 50 year status quo scenario in DCCS. This scenario demonstrated what the state of affairs would be if no action had been taken at all in the simulation. The scenario started in the year 2000 and ended in year 2050. In this scenario, the year 2000 emissions for deforestation were maintained at 1.3 GtC per year and the emissions for fossil fuels at 6.88 GtC per year for the next 50 years. For people put on the Above and Below conditions the initial concentration was kept at 1107 GtC and 769 GtC in the 50 year status quo scenario. If we start with year 2000 1
Recorded at the Mauna Loa observatory, Hawaii: http://www.esrl.noaa.gov/gmd/ccgg/trends/
6
CO2 concentration of 769 GtC i.e. Below condition, the concentration of CO2 in the atmospheric tank increased to more than 1000 GtC by year 2050, which is more than a 5% increase from the year 2000 value (this can be derived from the climate model detailed earlier). Participants were told of the severe consequences that this increase in concentration might hold in terms of economic and life losses under all the conditions of the experiment. After the scenario finished, participants were encouraged to ask questions on the task. Participants were given full information on the DCCS task and were reminded that their task was to control CO2 concentration to within the +/- 15 GtC range around the goal over the entire course of the DCCS task. They were then asked to play the DCCS task for 50 decision points over a course of 200 years either starting in Above or Below the goal conditions. Participants sat in front of a computer screen and were presented with the DCCS displayed in the center of the screen. We made use of Windows XP-based desktop computer terminals with plasma screens, which had a resolution of 1024 x 1080; the desktop computers ran on an Intel® Pentium 4 processor base. Optimal solution to climate change This section details optimal solutions for the two climate change conditions Above and Below using the dynamics we described for rate of CO2 transfer, frequency of emissions decision and other parameters in the preceding sections. These optimal solutions were developed using two different starting point of CO2 concentrations in the climate model detailed earlier. The model captured the basic elements of a climate system i.e. emissions much higher than absorptions, a positive relationship between the CO2 emissions and CO2 concentration, and tight restrictions on current CO2 emission reductions. It must be mentioned that although the values of CO2 concentration has been chosen in Above and Below conditions such that the CO2 concentrations are symmetrically around a realistic 938 GtC (~450 ppmv) goal (IPCC, 2001) these values actually create different task difficulty as described next specific to each condition of the experiment. Below Condition In the Below condition, as shown in Figure 3, the starting CO2 concentration is below an attainable goal of 938 GtC. The optimal solution (those that control the system in the minimum number of decision points) is obtained by increasing the fossil fuel and deforestation emissions at a maximum rate i.e. by maintaining their values at the To value ranges till the concentration approaches the goal. Once the concentration reaches goal the fossil fuel and deforestation emissions are reduced at a maximum rate i.e. by maintaining their values at the From range values till they equalize the CO2 absorptions (to stabilize CO2 concentration at the goal). It takes about 5 decision points to reach the goal range between 923 GtC and 953 GtC in Below condition. This will correspond to 20 (=5*4) years in DCCS.
7
Fig. 3. Optimal Total CO2 Emissions, Absorptions and CO2 concentration in the Below condition Above Condition In the Above condition, shown in Figure 4, the starting CO2 concentration is above an attainable goal of 938 GtC. The optimal solution is obtained by firstly decreasing the fossil fuel and deforestation emissions at a maximum rate i.e. by maintaining their values at the From range value till the concentration approaches the goal (emissions are below the absorptions and hence the concentration diminishes). Once the concentration reaches goal, the fossil fuel and deforestation emissions are increased at a maximum rate i.e. by maintaining their values at the To values of their respective ranges till they equalize the CO2 absorptions (this process will stabilize CO2 concentration at the goal). It takes about 18 decision points to reach the goal range between 923 GtC and 953 GtC in the Above condition. This will correspond to 72 (=18*4) years in DCCS.
8
Fig. 4. Optimal Total CO2 Emissions, Absorptions and CO2 concentration in the Above condition Thus, it becomes clear that as per the computed optimal strategies, the Below condition takes less number of decision points and hence it is easier than the Above condition in task difficulty where the Above condition takes more number of decision points to reach the goal. The next section details results of participant CO2 emissions decisions compared to optimal strategies reported in this section. Results Figure 5 and Figure 6 show the fits of optimal emissions derived in the last section to average human behavior shown for the resulting CO2 concentration, total emissions and absorptions obtained from our experiment in Below and Above conditions respectively. It can be seen from Figure 5 and Figure 6 that the participants did well to control the CO2 concentration to goal when they started Below than when they started Above (higher R2 and lower RMSD for Below condition than Above condition). Also, after comparing the average total CO2 emissions and average CO2 absorptions in Figure 5, we see that participants come close to the optimal
9
trajectories when starting below the goal. Whereas, as reported in Figure 6, the participant trajectories for average total CO2 emissions and average CO2 absorptions are similar in trend or shape to the optimal trajectories (R2 ~ 0.75) but they are highly deviated from the optimal trajectories (very high RMSD values). Also the trend in Above condition is not as good as the trend for average total CO2 emissions and average CO2 absorptions in the Below condition. These results show that participants average total CO2 emissions decisions were better when they started below the goal than when they started above. Also, similar comparisons can be made for average CO2 concentration, and average CO2 absorptions where these measures had a better fit to optimal for the Below condition than the Above condition.
10
Fig. 5. CO2 concentration, total emissions and absorptions from optimal solution and humans in the Below condition across 50 decision points (human curves are averaged over all participants in Below condition for each decision point). Error bars show 90% confidence interval around the point estimate. The R2 and RMSD are reported in each figure for fits of CO2 concentration, total emissions and absorptions of optimal solution to humans.
11
12
Fig. 6. CO2 concentration, total emissions and absorptions from optimal solution and humans in the Above condition across 50 decision points (human curves are averaged over all participants in Above condition for each decision point). Error bars show 90% confidence interval around the point estimate. The R2 and RMSD are reported in each figure for fits of CO2 concentration, total emissions and absorptions of optimal solution to humans. Time to Reach the Goal Here we analyze the number of decision points on average that participants took to reach the goal for the first time. The analysis of decision point to first reach the goal indicated that participants in the Below condition reached the goal faster (Mean = 16.46 decision points; SE = 3.47 decision points) than participants in the Above condition (Mean = 27.75 decision points; SE = 11.98 decision points). Figure 7 reports the comparison of Above and Below conditions for humans and optimal solution. Humans in both Above and Below conditions are far away from the optimal solutions in the number of decision points to reach the goal. The humans take less number of decision points to get CO2 concentration to reach the goal range for Below as compared to Above condition and likewise is the case for the optimal solutions for the two conditions.
Fig. 7. Average Decision Point to reach the goal under Above and Below conditions for humans and optimal solutions. The results show that the Humans were away from the optimal by about 11 Decision Points across the Above and Below Conditions (RMSD ~ 11). Error bars show 90% confidence interval around the point estimate. Discussion and conclusions Many of the complex dynamic effects found in the real world (like CO2 dynamics) can be better understood with simple tasks (Cronin et al., 2009) and here we present a demonstration of such a process in DCSS. The results reported in this paper clearly support an argument in favor of our initial expectation of poorer control performance in a system where participants decrease their
13
emissions in order to reduce concentration towards an attainable goal, than in a system where they increase their emissions in order to increase concentration towards the same goal. Our results using different dependent measures indicate that participants starting above the goal exhibited poor control performance than those starting below the goal (lower R2 and higher RMSD values for humans when compared with optimal for Above than Below condition). Our results also reveal that the participant control performance as seen in their CO2 concentration’s control to goal in DCCS generally remained poor although the control was better and similar to the optimal solution in the Below condition than the Above condition. Further, participants made poor selection of average total emissions decisions in the Above than in the Below condition when both conditions were compared with the optimal solution. The deterioration of control performance for starting above the goal than below is due to one simple reason: task difficulty on account of basic climatic factors where emissions are greater than absorptions, emissions are restricted to upper and lower bounds and emissions are the only decision making elements directly controlled by the decision makers. Research Implications This research provides interesting cognitive explanations for real world climate policy making that involve reducing emissions as per the Kyoto Protocol and those of IPCC’s higher than current concentration goal setting practices. Because people find it harder to understand, perceive and control the system dynamics for our climate when they are above the goal than when they are below, on account of task’s difficulty and their cognitive limitations, they prefer to set goals that are higher relative to where they stand today. Although, there could be many economic, social and political reasons yet one simple cognitive reason explains why Kyoto has been a failure and why IPCC sets goals that are higher relative to where we stand today – people’s tendency to easily perceive and prefer challenges that are less difficult than those that are more difficult to attain. The main policy implication from this research is that there is currently some bad news in terms of future of policy making for our climate from a purely cognitive view point. People want to anchor and adjust from their current positions relative to where they stand but prefer to do it in a direction that provides them a path to lessen the problem’s difficulty i.e. where their deficient mental representations are more easily satisficed (Simon, 1990; Tversky, and Kahneman, 1974). This anchor and adjustment and satisficing behavior for a dynamic system like the climate from our current research results means that people will have a constant tendency to look upwards and set higher goals as the world attains even higher levels for the rising atmospheric CO2 concentrations in the future. This behavior on part of humans will prove extremely detrimental for our climate. Today we are crossing the 800 GtC CO2 concentration mark and still going up (as per the Mauna Loa 2009 estimates2) and because we are cognitively able to think upwards easily than downwards, we are sadly thinking of making higher stabilization plans which are in excess of 1000 GtC today. We plan to continue our initial research investigations reported in this paper. In particular, it is important to understand how the dynamic complexity posed by the climate dynamics (which was assumed fixed in this research) would interact with the task difficulty of different initial or starting CO2 concentrations. For example, it would be important to see how rapid dynamics, where the climate removes CO2 more rapidly from the atmosphere (i.e. at a higher Rate of CO2 transfer), helps people controlling the climate system better even when people start above the 2
http://www.esrl.noaa.gov/gmd/ccgg/trends/
14
goal. As for another follow up experiment, it would be good to determine if our current hypothesis on task difficulty will become true for situations where people start at the same initial value of CO2 concentration in all conditions. But now have to control CO2 concentration to two symmetric and attainable goals from the common starting point, one of which is above and the other below the common starting point. We see that greater understanding of human incapacities and cognitive limitations using tools like DCCS is the right step for research in this endeavor. Acknowledgments This research was partially supported by the National Science Foundation (Human and Social Dynamics: Decision, Risk, and Uncertainty, Award number: 0624228) award to Cleotilde Gonzalez.
15
REFERENCES Bryner G. 2007. Lessons from the Kyoto Protocol for the Design of Future Global Climate Accords. In Proceedings, the annual meeting of the International Studies Association 48th Annual Convention. Chicago, IL. Cronin M, Gonzalez C, Sterman JD. 2009. Why don't well-educated adults understand accumulation? A challenge to researchers, educators and citizens. Journal of Organizational Behavior and Human Decision Processes 108(1): 116-130. Dutt V, Gonzalez C. 2008. Human Perceptions of Climate Change. In Proceedings, 26th International Conference of the System Dynamics Society. Athens, Greece. [CD-ROM] IPCC. 2001. Climate Change 2001: The Scientific Basis. Intergovernmental Panel on Climate Change 83. IPCC. 2007. Climate Change 2007: Synthesis Report. Contribution of Working Groups I, II and III to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change [Core Writing Team, Pachauri, R.K and Reisinger, A. (eds.)]. IPCC. Geneva, Switzerland, pp. 104. Jain AK, Kheshgi HS, and Wuebbles DJ. 1994. Integrated Science Model for Assessment of Climate Change. In Proceedings, Air and Waste Management Association’s 87th Annual Meeting. Cincinnati, Ohio. Prins G, Rayner S. 2007. Time to ditch Kyoto. Nature 449: 973-975. Simon HA. 1990. Invariants of Human Behavior. Annual Review of Psychology 41: 1-19. Sterman JD, Booth Sweeney L. 2002. Cloudy skies: Assessing public understanding of global warming. System Dynamics Review 18(2): 207-240. Sterman JD, Booth Sweeney L. 2007. Understanding public complacency about climate change: Adults' mental models of climate change violate conservation of matter. Climatic Change 80(3-4): 213-238. Tversky A, Kahneman D. 1974. Judgment under uncertainty: Heuristics and biases. Science 185(4157): 1124-1131. United Nations. 1998. Framework Convention on Climate Change. Kyoto Protocol. Climate Change Secretariat (UNFCCC). HTTP: http://unfccc.int/kyoto_protocol/items/2830.php (Retrieved on 17th March 2009). [Online]. Victor DG. 2004. The Collapse of the Kyoto Protocol and the Struggle to Slow Global Warming (Council on Foreign Relations Book). Princeton, MA: Princeton University Press.
16
