Less Wrong is a community blog devoted to refining the art of human rationality. Please visit our About page for more information.
Alright guys. The main complaint of the discussion article was simply "hoax", yelled as loudly or as quietly as the user felt about it. Hopefully this won't get the same treatment.
We have been evaluating educational, grant-funded programs for 20 years. Throughout these years, we have witnessed a slow change in how students are selected for academic services. Traditionally, students were targeted for academic services and opportunities based on demographic characteristics—usually race and, until recently, family income status (based on free or reduced priced lunch). Wealthier, white students are given challenging lessons and tracked into the advanced courses, while their non-white and poorer peers are tracked low and given remediation services. The latter students are often referred to as “at-risk,” though we are finding more and more that the greatest risk these students face is being placed into inappropriate remedial courses which eventually bar them from access to advanced courses. After students have been labeled “at-risk,” and then tracked inappropriately and provided unnecessary (and often harmful) remediation, their downward trajectory continues throughout their education. The demographic gap this creates continues to expand, despite the lip service and excessive tax and grant funds paid to eliminate—or at least lessen—this very gap. This “at-risk” model of assigning services is slowly being replaced by a “pro-equity” model. The driving force behind this change is the availability and use of data.
The literature is full of documentation that certain demographic groups have traditionally had less access to advanced math and science courses than equally scoring students belonging to demographic groups thought to be “not at risk.” Some examples from research follow.
• Sixth grade course placement is the main predictor of eighth grade course placement, and social factors--mainly race---are key predictors of sixth grade course placement (O’Connor, Lewis, & Mueller, 2007).
• Among low-income students, little is done to assess which are high achievers. Few programs are aimed at them, and their numbers are lumped in with “adequate” achievers in No Child Left Behind reporting. As a result, little is known about effective practices for low-income students (Wyner, Bridgeland, & DiIulio Jr., 2007).
• In a California school district, researchers found that of students who demonstrated the ability to be admitted to algebra, 100% of the Asians, 88% of the whites, 51% of the Blacks, and 42% of the Latinos were admitted (Stone & Turba, 1999).
• Tracking has been described as “a backdoor device for sorting students by race and class.” Many researchers agree (Abu El-Haj & Rubin, 2009).
• When course grades are used to determine placement, studies show that some students’ grades “matter” more than others. Perceptions of race and social class are often used to determine placement (Mayer, 2008).
• Studies show that when schools allow students the freedom to choose which track they’ll take, teachers and counselors discourage many previously lower tracked students from choosing the higher track (Yonezawa, Wells, & Serna, 2002).
• The sequence of math students take in middle school essentially determines their math track for high school. In North Carolina, this is true because of math prerequisites for higher level math (North Carolina Department of Public Instruction, 2009).
We are seeing a move toward using objective data for placement into gateway courses, such as 8th grade algebra. Many school districts are beginning to use Education Value Added Assessment (EVAAS) and other data system scores that predict success in 8th grade algebra for criteria to enroll. This pro-equity model is replacing the traditional, at-risk model that relied on professional judgment. One example of this is in Wake County, North Carolina. Superintendent Tony Tata attributed a 44% increase in the number of students enrolled in algebra to the use of the predictive software, EVAAS, to identify students likely to be successful. The success rate in the course increased with the addition of these students (KeungHu, 2012).
Although the pro-equity model of using objective data to assign students to more rigorous courses has proven successful, many people resist it. These people cling to the at-risk model, dismissing the objective data as inconclusive. Many of the overlooked students who were predicted to succeed, yet were placed in lower tracks (disproportionately minorities), are “weaker,” according to the old-school staff, and allowing these students into the gateway 8th-grade algebra course would be a disservice to them. (Not allowing them into this course ensures their bleak academic future.) Review of the data had shown that strong students were being overlooked, and this objective use of data helps identify them (Sanders, Rivers, Enck, Leandro, & White, 2009).
The changes in education began with concern for aligning academic services with academic need. Aligning opportunities for rigor and enrichment is only just beginning. In the past, a large proportion of federal grant funds were for raising proficiency rates. In the at-risk model, grant funds were provided for services to the minority and poor demographic groups with the goals of raising academic proficiency rates. When we first started evaluating grant-funded programs, most federal grants were entirely in the at-risk model. The students were targeted for services based on demographic characteristics. The goals were to deliver the services to this group. Staff development was often designed to help staff understand children in poverty and what their lives are like, rather than helping them learn how to deliver an effective reading or math intervention. The accountability reports we were hired to write consisted of documentation that the correct demographic group was served, the program was delivered, and staff received their professional development. Proficiency rates were rarely a concern.
In 2004, the federal government developed the Program Assessment Rating Tool (PART) to provide accountability to grant-funded programs by rating their effectiveness. The PART system assigned scores to programs based on services being related to goals, showing that the goals were appropriate for the individuals served, and student success measured against quality standards and assessments. PART rated programs that could not demonstrate whether they have been effective or not because of lack of data or clear performance goals with the rating “Results Not Demonstrated” (U.S. Office of Management and Budget and Federal Agencies, n.d. "The Program Assessment Rating Tool") . In 2009, nearly half (47%) of U.S. Department of Education grant programs rated by the government are given this rating, thus illustrating the difficulties of making this transition to outcome based accountability (U.S. Office of Management and Budget and Federal agencies, n.d. "Department of Education programs"). The earliest changes were in accountability, not in program services or how to target students. Accountability reports began asking for pre- and post-comparisons of academic scores. For example, if funds were for raising the proficiency rates in reading, then evaluation reports were required to compare pre- and post-reading scores. This was a confusing period, because programs still targeted students based on demographic information and provided services that often had no research basis linking them to academic achievement; professional development often remained focused on empathizing with children in poverty, although the goals and objectives would now be written in terms of the participants raising their academic achievement to proficiency. We evaluators were often called in at the conclusion of programs to compare pre- and post-academic scores, and determine whether participants improved their scores to grade-level proficiency. We often saw the results of capable students treated like low-achievers, thought to have no self-esteem, and provided remedial work. Such treatment damaged the participants who had previously scored at or above proficient prior to services.
A typical narrative of an evaluation might read:
The goal of the program was to raise the percentage of students scoring proficient in reading. The program targeted and served low-income and minority students. Staff received professional development on understanding poor children. Services offered to students included remedial tutorials and esteem-building activities. When the program ended, pre-reading scores were obtained and compared with post-scores to measure progress toward the program objective. At that time, it was discovered that a large percentage of participants were proficient prior to receiving services.
Rather than cite our own evaluations, we found many examples from the school districts reporting on themselves.
Accelerated Learning Program.
The following is a direct quote from a school system in North Carolina:
. . . Although ALP [Accelerated Learning Program] was designed primarily to help students reach proficiency as measured by End-of-Grade (EOG) tests, only 41.1% of those served showed below-grade-level scores on standard tests before service in literacy. In mathematics, 73.3% of students served had below-grade-level scores. ALP served about 40% of students who scored below grade level within literacy and within mathematics, with other services supporting many others. . . . Compared to those not served, results for Level I-II students were similar, but results for Level III-IV students were less positive. One third of non- proficient ALP mathematics students reached proficiency in 2008, compared to 42.1% of other students. (Lougee & Baenen, 2009).
Foundations of Algebra
This program was designed for students who fit specific criteria, yet it served many students who did not. Students who were below proficient or almost proficient were to be placed in courses to eventually prepare them for Algebra I. When criteria for placement are not met, determining program effectiveness is difficult, if not impossible. Students were likely entered into the program based on teacher recommendations, which were subsequently based on demographic factors such as race. The teachers “mistook” these students for below-proficient students when they were not. Had objective data, such as actual proficiency scores, been consulted, the proper students could have been served. The report indicates a success, as a higher percentage of these students than similar students who were not served enrolled in Algebra I. However, it is not known if this comparison group includes only students who actually meet the criteria, or if they are a heterogeneous mix of students of varying abilities. Missing data also makes program effectiveness evaluation difficult (Paeplow, 2010).
Partnership for Educational Success (PES)
This program was purportedly for students who are “at risk,” which is defined as students who scored below grade level on EOG (below proficiency) and have been “identified by the PES team as having family issues that interfere with school success.” What is meant by “family issues” is unclear. The majority of students served are Economically Disadvantaged (ED) (91.3%) and Black (71.5%). More than half the students served, according to the evaluation, were at or above grade level on their EOGs when they began the program, thus making program effectiveness difficult to judge. The family component is an integral part of the program, and outside agencies visit families. Many community organizations are involved. But if the staff could miss so easy a datum as EOG scores for so many students, one has to wonder about such a subjective criterion as “family issues.” The program appears to have targeted ED students, with little regard to prior performance data. Data for many students (43.5%) was missing. Teachers indicate that parents of the targeted families have become more involved in the school, but little else has changed (Harlow & Baenen, 2004).
Helping Hands was initiated based on data indicating that Black males lag behind other groups in academic achievement. The program is supposed to serve Black males, and most of the participants fit these criteria. The program is also designed to improve academics, and to curtail absenteeism and suspensions. Although the percentage of selected participants who needed improvement in these areas was higher than it was for the overall population of the students served, not all students served demonstrated a need for intervention. Many students were at grade level, were not chronically absent, and had not been suspended. Yet they were served because they were Black and male (Paeplow, 2009).
At Hodge Road Elementary School, students were tutored with remedial work in an after-school program. The only criterion the students had to meet to be allowed into the program was the inability to pay full price for their lunch. Their academic performance was irrelevant. (To be fair, these criteria were instituted by No Child Left Behind, and not the school system.) Most students were already reading and doing math at or above grade level (the two subjects for which tutoring was provided). The evaluation shows that giving remedial coursework to students who are at or above grade level, as if they were below grade level, can actually harm them. In the final statistics, 11.1% of Level III & IV 3rd through 5th graders scored below grade level after being served, compared with only 2% of a comparable group who were not served. An astonishing 23% of students in kindergarten through 2nd grade served who were at or above grade level prior to the tutoring scored below grade level afterward, compared with 8% of comparable students who were not served (Paeplow & Baenen, 2006).
AVID is a program designed for students who may be the first in their families to attend college, and who are average academic performers. The program, developed in the 1980s, maintains that by providing support while holding students to high academic standards, the achievement gap will narrow as students succeed academically and go on to successfully complete higher level education. Fidelity of implementation is often violated, which, as proponents admit on AVID’s own website (www.AVID.org) may compromise the entire program. Student participants must have a GPA of 2.0-3.5. We were asked to evaluate Wake County Public School Systems AVID program. Many students chosen for the program, however, did not fit the criteria (Lougee & Baenen, 2008). Because AVID requirements were not met, a meaningful evaluation was not possible.
This AVID program was implemented with the goal of increasing the number of under-represented students in 8th grade algebra. This was at a time when no criteria for enrollment in 8th grade algebra existed (i.e., a target to help the students reach didn’t exist), and high scoring students in this very group were not being referred for enrollment in algebra. Under these conditions, the program makes no sense. In summary, the goal of this program is to enroll in 8th grade algebra more low-income, minority, and students whose parents didn’t go to college. Only students recommended by teachers can enroll in 8th grade algebra. The data showed that very high-scoring, low-income and minority students were not being recommended for 8th grade algebra. Why do we think that students whose parents didn’t go to college can’t enroll in 8th grade algebra without being in an intervention program first? (Also, how it is determined that the students’ parents did not attend college is not addressed.) The program is for low-average students. They served high-average students. Then they still didn’t recommend them to be in 8th grade algebra. This program is very expensive. We have evaluated this program in many school districts and we find the same results, typically, as this report.
During this era, the interventions typically have not been related to the desired outcomes by research. For example, self-esteem-building activities were often provided to increase the odds of passing a math class, or to improve reading scores. Sometimes programs would be academic, but claims for success were not research-based, nor was the relationship between the activities and the desired outcomes. Although many interventions were at least related to the academic subject area the program was trying to impact, it was not unheard of to see relaxation courses alone for increasing math test scores, or make-overs and glamor shots for raising self-esteem, which in turn would allegedly raise reading scores.
During the last decade, education has slowly moved toward requiring accountability in terms of comparing pre- and post-scores. We saw this causing confusion and fear, rather than clarity. More than once, when we reported to school districts that they had served significant numbers of students who were already at or above proficiency levels, they thought we were saying they had served high-income students instead of their target population of low-income students. We have seen many school systems assess their own programs, write evaluation reports like the examples above, and then continue to implement the programs without any changes. We have worked with some educators whose eyes were opened to the misalignment of services and needs, and they learned to use data, to identify appropriate interventions, and keep records to make accountability possible. We’ve seen these innovators close their achievement gaps while raising achievement of the top. But, those around them didn’t see this as replicable.
Race to the Top will impact the rate of change from the at-risk to the pro-equity model. Teacher and principal evaluations are going to include measures of growth in student learning (White House Office of the Press Secretary, 2009). EVAAS will be used to measure predicted scores with observed scores. If high-achieving students who are predicted to succeed in 8th grade algebra are tracked into the less rigorous 9th grade algebra, they are not likely to make their predicted growth .
We are moving out of this era, and the pace of change toward identifying student needs using appropriate data is picking up. North Carolina’s newly legislated program, Read to Achieve, mandates that reading interventions for students in K-3 be aligned to the literacy skills the students struggle with, and that data be used to determine whether students are struggling with literacy skills. Schools must also keep records for accountability. Although this approach seems logical, it is quite innovative compared with the past reading interventions that targeted the wrong students (North Carolina State Board of Education; Department of Public Instruction, n.d.).
Education Grant programs are now requiring that applicants specify what data they will use to identify their target population, and how the intervention relates to helping the participants achieve the program goals. Staff development must relate to delivering the services well, and accountability must show that these things all happened correctly, while documenting progress toward the program objectives. It is a new era. We are not there yet, but it is coming.
Harlow, K., & Baenen, N. (2004). E & R Report No. 04.09: Partnership for Educational Success 2002-03: Implementation and outcomes. Raleigh, NC: Wake County Public School System. Retrieved from http://www.wcpss.net/evaluation-research/reports/2004/0409partnership_edu.pdf
KeungHu. (2012). Wake County Superintendent Tony Tata on gains in Algebra I enrollment and proficiency. Retrieved from http://blogs.newsobserver.com/wakeed/wake-county-superintendent-tony-tata-on-gains-in-algebra-i-enrollment-and-proficiency
Lougee, A., & Baenen, N. (2008). E & R Report No. 08.07: Advancement Via Individual Determination (AVID): WCPSS Program Evaluation. Retrieved from http://www.wcpss.net/evaluation-research/reports/2008/0807avid.pdf
Lougee, A., & Baenen, N. (2009). E&R Report No. 09.27: Accelerated Learning Program (ALP) grades 3-5: Evaluation 2007-08. Retrieved from http://www.wcpss.net/evaluation-research/reports/2009/0927alp3-5_2008.pdf
Mayer, A. (2008). Understanding how U.S. secondary schools sort students for instructional purposes: Are all students being served equally? . American Secondary Education , 36(2), 7–25.
North Carolina Department of Public Instruction. (2009). Course and credit requirements. Retrieved from http://www.ncpublicschools.org/curriculum/graduation
North Carolina State Board of Education; Department of Public Instruction. (n.d.). North Carolina Read to Achieve: A guide to implementing House Bill 950/S.L. 2012-142 Section 7A. Retrieved from https://eboard.eboardsolutions.com/Meetings/Attachment.aspx?S=10399&AID=11774&MID=783
O’Connor, C., Lewis, A., & Mueller, J. (2007). Researching “Black” educational experiences and outcomes: Theoretical and methodological considerations. Educational Researcher. Retrieved from http://www.sociology.emory.edu/downloads/O%5c’Connor_Lewis_Mueller_2007_Researching_black_educational_experiences_and_outcomes_theoretical_and_methodological_considerations.pdf
Paeplow, C. (2009). E & R Report No. 09.30: Intervention months grades 6-8: Elective results 2008-09. Raleigh, NC: Wake County Public School System. Retrieved from http://www.wcpss.net/evaluation-research/reports/2009/0930imonths6-8.pdf
Paeplow, C. (2010). E & R Report No. 10.28: Foundations of Algebra: 2009-10. Raleigh, NC: Wake County Public School System. Retrieved from http://assignment.wcpss.net/results/reports/2011/1028foa2010.pdf
Paeplow, C., & Baenen, N. (2006). E & R Report No. 06.09: Evaluation of Supplemental Educational Services at Hodge Road Elementary School 2005-06. Raleigh. Retrieved from http://www.wcpss.net/evaluation-research/reports/2006/0609ses_hodge.pdf
Sanders, W. L., Rivers, J. C., Enck, S., Leandro, J. G., & White, J. (2009). Educational Policy Brief: SAS® Response to the “WCPSS E & R Comparison of SAS © EVAAS © Results and WCPSS Effectiveness Index Results,” Research Watch, E&R Report No. 09.11, March 2009. Cary, NC: SAS. Retrieved from http://content.news14.com/pdf/sas_report.pdf
Stone, C. B., & Turba, R. (1999). School counselors using technology for advocacy. Journal of Technology in Counseling. Retrieved from http://jtc.colstate.edu/vol1_1/advocacy.htm
U.S. Office of Management and Budget and Federal Agencies. (n.d.). The Program Assessment Rating Tool (PART). Retrieved from http://www.whitehouse.gov/omb/expectmore/part.html
U.S. Office of Management and Budget and Federal agencies. (n.d.). Department of Education programs. Retrieved from http://www.whitehouse.gov/omb/expectmore/agency/018.html
White House Office of the Press Secretary. (2009). Fact Sheet: The Race to the Top. Washington D.C. Retrieved from http://www.whitehouse.gov/the-press-office/fact-sheet-race-top
Wyner, J. S., Bridgeland, J. M., & DiIulio Jr., J. J. (2007). Achievement trap: How America is failing millions of high-achieving students from low-income families. Jack Kent Cooke Foundation, Civic Enterprises, LLC. Retrieved from www.jkcf.org/assets/files/0000/0084/Achievement_Trap.pdf
Yonezawa, S., Wells, A. S., & Serna, I. (2002). Choosing tracks:“Freedom of choice” in detracking schools. American Educational Research Journal , 39(1), 37–67.
tl;dr: Playing the true PD, it might be that you should co-operate when expecting the other one to defect, or vice versa, in some situations, against agents that are capable of superrationality. This is because relative weight of outcomes for both parties might vary. This could lead this sort of agents to outperform even superrational ones.
So, it happens that our benevolent Omega has actually an evil twin, that is as trustworthy as his sibling, but abducts people into a lot worse hypothetical scenarios. Here we have one:
You wake up in a strange dimension, and this Evil-Omega is smiling at you, and explains that you're about to play a game with unknown paperclip maximizer from another dimension that you haven't interacted with before and won't interact with ever after. The alien is like GLUT when it comes to consciousness, it runs a simple approximation of rational decision algorithm but nothing that you could think of as "personality" or "soul". Also, since it doesn't have a soul, you have absolutely no reason to feel bad for it's losses. This is true PD.
You are also told some specifics about the algorithm that the alien uses to reach its decision, and likewise told that alien is told about as much about you. At this point I don't want to nail the algorithm the opposing alien uses down to one specific. We're looking for a method that wins when summing up all these possibilities. Next, especially, we're looking at the group of AI's that are capable of superrationality, since against other's the game is trivial.
The payoff matrix is like this:
DD=(lose 3 billion lives and be tortured, lose 4 paperclips), CC=(2 billion lives and be made miserable, lose 2 paperclips), CD=(lose 5 billion lives and be tortured a lot, nothing), DC=(nothing, lose 8 paperclips)
So, what do you do? Opponent is capable of superrationality. In the post "The True Prisoner's Dilemma", it was(kinda, vaguely, implicitly) assumed for simplicity's sake that this information is enough to decide whether to defect or not. Answer, based on this information, could be to co-operate. However, I argue that information given is not enough.
Back to the hypothetical: In-hypothetical you is still wondering about his/her decision, but we zoom out and observe that, unbeknownst to you, Omega has abducted your fellow LW reader and another paperclip maximizer from that same dimension, and is making them play PD. But this time their payoff matrix is like this:
DD=(lose $0.04, make 2 random, small changes to alien's utility function and 200 paperclips lost), CC=(lose $0.02, 1 change, 100 paperclips), CD=(lose $0.08, nothing), DC=(nothing, 4 changes, 400 paperclips)
Now, if it's not "rational" to take the relative loss into account, we're bound to find ourselves in a situation where billions of humans die. You could be regretting your rationality, even. It should become obvious now that you'd wish you could somehow negotiate both of these PD's so that you would defect and your opponent co-operate. You'd be totally willing to take a $0.08 hit for that, maybe paying it in its entirety for your friend. And so it happens, paperclip maximizers would also have an incentive to do this.
But, of course, players don't know about this entire situation, so they might not be able to operate in optimal way in this specific scenario. However, if they take into account how much the other cares about those results, using some unknown method, they just might be able to systematically perform better(if we made more of this sorts of problems, or if we selected payoffs at random for the one-shot game), than "naive" PD-players playing against each other. Naivity here would imply that they simply and blindly co-operate against equally rational opponents. How to achieve that is the open question.
Stuart Armstrong, for example, has an actual idea of how to co-operate when the payoffs are skewed, while I'm just pointing out that there's a problem to be solved, so this is not really news or anything. Anyways, I still think that this topic has not been explored as much as it should be.
Edit. Added this bit: You are also told some specifics about the algorithm that the alien uses to reach its decision, and likewise told that alien is told about as much about you. At this point I don't want to nail the algorithm the opposing alien uses down to one specific. We're looking for a method that wins when summing up all these possibilities. Next, especially, we're looking at the group of AI's that are capable of superrationality, since against other sort of agents the game is trivial.
Edit. Corrected some huge errors here and there, like, mixing hypothetical you and hypothetical LW-friend.
Edit. Transfer Discussion -> Real LW complete!
[His message: Neurological evidence suggests - somewhat alarmingly - that our moral and ethical decisions may be no more than post-hoc rationalisations of purely emotional, instinctive reactions. However, we should not panic because this is early days in neuroscience, and the correct interpretation of brain-scans is uncertain: scientist find the pattern, and the explanation, they expect to find]
We're interested in improving human rationality. Many of our techniques for improving human rationality take time. In real-time situations, you can lose by making the wrong decision, or by making the "right" decision too slowly. Most of us do not have inflexible-schedule, high-stakes decisions to make, though. How often does real-time decision making really come up?
Suppose you are making a fairly long-ranged decision. Call this decision 1. While analyzing decision 1, you come to a natural pause. At this pause you need to decide whether to analyze further, or to act on your best-so-far analysis. Call this decision 2. Note that decision 2 is made under tighter time pressure than decision 1. This scenario argues that decision-making is recursive, and so if there are any time bounds, then many decisions will need to be made at very tight time bounds.
A second, "covert" goal of this post is to provide a definitely-not-paradoxical problem for people to practice their Bayseian reasoning on. Here is a concrete model of real-time decisionmaking, motivated by medical-drama television shows, where the team diagnoses and treats a patient over the course of each episode. Diagnosing and treating a patient who is dying of an unknown disease is a colorful example of real-time decisionmaking.
To play this game, you need a coin, two six-sided dice, a deck of cards, and a helper to manipulate these objects. The manipulator sets up the game by flipping a coin. If heads (tails) the patient is suffering from an exotic fungus (allergy). Then the manipulator prepares a deck by removing all of the clubs (diamonds) so that the deck is a red-biased (black-biased) random-color generator. Finally, the manipulator determines the patients starting health by rolling the dice and summing them. All of this is done secretly.
Play proceeds in turns. At the beginning of each turn, the manipulator flips a coin to determine whether test results are available. If test results are available, the manipulator draws a card from the deck and reports its color. A red (black) card gives you suggestive evidence that the patient is suffering from a fungus (allergy). You choose whether to treat a fungus, allergy, or wait. If you treat correctly, the manipulator leaves the patient's health where it is (they're improving, but on a longer timescale). If you wait, the manipulator reduces the patient's health by one. If you treat incorrectly, the manipulator reduces the patient's health by two.
Play ends when you treat the patient for the same disease for six consecutive turns or when the patient reaches zero health.
Here is some Python code simulating a simplistic strategy. What Bayesian strategy yields the best results? Is there a concise description of this strategy?
The model can be made more complicated. The space of possible actions is small. There is no choice of what to investigate next. In the real world, there are likely to be diminishing returns to further tests or further analysis. There could be uncertainty about how much time pressure there is. There could be uncertainty about how much information future tests will reveal. Every complication will make the task of computing the best strategy more difficult.
We need fast approximations to rationality (even quite bad approximations, if they're fast enough), as well as procedures that spend time in order to purchase a better result.
While we have all of us here together to crunch on problems, let's shoot higher than trying to think of solutions and then finding problems that match the solution. What things are unsolved questions? Is it reasonable to assume those questions have concrete, absolute answers?
The catch is that these problems cannot be inherently fuzzy problems. "How do I become less wrong?" is not a problem that can be clearly defined. As such, it does not have a concrete, absolute answer. Does Rationality have a set of problems that can be clearly defined? If not, how do we work toward getting our problems clearly defined?
See also: Open problems at LW:Wiki
View more: Next