Reevaluating the effectiveness of n-back training on transfer through the Bayesian lens: Support for the null.
445 views
Skip to first unread message
XFMQ902SF
Jun 18, 2015, 10:22:08 PM6/18/15
to brain-t...@googlegroups.com
A recent meta-analysis by Au et al. Psychonomic Bulletin & Review, 22, 366-377, (2015) reviewed the n-back training paradigm for working memory(WM) and evaluated whether (when aggregating across existing studies) there was evidence that gains obtained for training tasks transferred to gains in fluid intelligence (Gf). Their results revealed an overall effect size of g = 0.24 for the effect of n-back training on Gf. We reexamine the data through a Bayesian lens, to evaluate the relative strength of the evidence for the alternative versus null hypotheses, contingent on the type of control condition used. We find that studies using a noncontact (passive) control group strongly favor the alternative hypothesis that training leads to transfer but that studies using active-control groups show modest evidence in favor of the null. We discuss these findings in the context of placebo effects.
Gwern Branwen
Jun 19, 2015, 12:44:43 PM6/19/15
to N-back
Fulltext: https://www.dropbox.com/s/lrqb9ajthk0o6tv/2015-dougherty.pdf
/ http://sci-hub.org/downloads/1b56/dougherty2015.pdf
> While we commend Au et al. on a rigorous meta-analysis, we contend that their analysis insufficiently address these issues. For example, while Au et al. relied on well-established null hypothesis significance testing (NHST) methods for meta-analysis, two well-known limitations of the NHST framework are that it tends to overstate evidence for the alternative hypothesis and does not permit one to evaluate the relative probability that the null hypothesis is in fact true. 1 In the context of the WM training literature, both of these problems are especially salient because the primary issue of debate is if working memory training is effective at all. This implies a need to evaluate the degree to which the data support the alternative hypothesis relative to the null, and is most easily addressed within a Bayesian approach.
Especially if you use an informative prior for how often intelligence
interventions fail to boost scores on IQ tests, and how often IQ test
score increases then fail to turn out to be on the latent g factor.
> Au et al. (2015) made an excellent attempt to reduce the potential influence of publication bias, with many studies included from nonpublished reports. The selection of studies to be included in the analysis appears to have been thorough and fair.
lol no.
> Au et al. (2015) presented effect sizes for 24 individual comparisons drawn from 20 papers. The aggregate weighted effect size across these 24 comparisons was 0.24. They also evaluated several possible mediators, including whether the studies used an active control (N = 12) or a passive control (N = 12), which yielded effect sizes of 0.06 and 0.44, respectively. Although Au et al. reported this effect as significant, they concluded that type of control group did not moderate the effect. This strikes us as an odd conclusion given that the magnitudes of these effect sizes differ considerably. Au et al.’s conclusion was based on a comparison between the control groups for active and passive studies, not by comparing the control groups to the treatment condition. The comparison of control groups while ignoring the training groups isn’t particularly informative regarding effect of training, since the effects of training can only be assessed relative to the control. In this regard, it is interesting to note that the effect size for the training condition amongst active-control studies (d = 0.25) is actually numerically smaller than the effect size amongst the control participants in the passive control studies (d = 0.28). The question is: Do these effect sizes provide evidence for training effectiveness?
Indeed. But in Au's defense, they claimed that an interaction with
non-American samples is driving the inflation of passive control
groups, not that the inflation exists (as it obviously does just
looking at the chart).
> The first step of our analysis involves transforming the effect sizes presented in Fig. 3 of Au et al. to their corresponding t values using t = sqrt(1/n1 + 1/n2) * g, where g is the measure of effect size and n1 and n2 are the sample sizes for two independent groups used in the effect-size calculations. We then computed the default Bayes factor (BF) corresponding to each t statistic using the ttestBF function in the BayesFactor package in R (Morey, Rouder, & Jamil, 2014; R Core Team, 2014) as well as the meta-analytic Bayes factor using the meta.ttestBF function. For all analyses, we set the scale factor on effect size to r = 1 and used a one-sided interval, which places the mass of the prior on effects greater than zero. The one-sided test is a reasonable assumption under the hypothesis that training should lead to improvements in Gf. Importantly, even large modifications to the prior distribution do not alter our conclusions in any substantive way, nor does using a two-sided null interval.
>
> ...As should be evident from Fig. 1 and Table 1, few of the individual studies provide particularly strong evidence for either the null or the alternative. Yet, looking across the entirety of the results, a curious pattern is obvious. First, 11 of the 12 effect sizes for the passive control studies are positive, whereas only 6 of the 12 effect sizes are positive for the active-control studies. Second, when these effect sizes are evaluated in terms of the Bayes factor, the majority of the individual studies favor the null hypothesis, including 6 of the 12 passive-control studies. These individual results using the BF roughly mirror the conclusions drawn from the significance tests, though the BF illustrates that the bulk of the studies show evidence for the null. However, these individual comparisons do not capitalize on a major strength of meta-analytic techniques, which is the ability to aggregate across studies to overcome the sample size problem. Moving on to the meta-analytic results, here the results diverge somewhat from the conclusions garnered from the individual studies. First, ignoring the type of control, the odds in favor of the alternative hypothesis is 152:1. This qualifies as 'decisive' evidence according to Jeffreys’ (1961) scheme. Figure 2, which provides the BFs conditioned on the use of passive- versus active-control groups, paints a much different picture. While the Bayes factor for the passive control studies is a whopping 13,241:1 in favor of the alternative, the Bayes factor for the active control studies is a more modest 7.7:1 in favor of the null.
A hypothesis-testing framework is ugly though, and makes it harder to
examine Au et al's international claim, so:
> Thus, we conducted a series of follow-up analyses using hierarchical Bayesian modeling, in which we modeled the effect sizes as a function of control group type (passive vs. active) as well as an additive effect of both control group type and country of origin (USA vs. non- USA). Au et al. (2015) identified country of origin as an im- portant moderator variable, with studies conducted within the USA yielding a small nonsignificant effect size and studies conducted outside the USA resulting in a moderate significant effect size – an effect that Au et al. hypothesized could be due to differences in motivation or compliance between USA and non-USA subjects. The inclusion of country of origin in our analysis allowed us to control for a potential important source of variability that Au et al. (2015) felt was theoretically justified. As we illustrate, inclusion of this variable in the Bayesian model reveals that the only estimated effect sizes that are different from zero are those based on non-USA passive-control studies. Furthermore, the estimated effect size for the active-control studies within the USA shrink to essentially zero...Strikingly, even when the prior distribution is set such that the effect of training is assumed to be large, there is still no evidence of that n-back training leads to improvements on Gf measures. While this model estimates that the median effect size amongst the active control studies is slightly above zero, this small positive effect is essentially eliminated when country of origin is added as a predictor in the model, as shown in Figs. 5 and 6. Importantly, the three international studies using active controls fail to yield a reliable positive effect. Furthermore, at the aggregate level the only effect size in which the HDI does not include zero are effects based on studies conducted outside the USA that use passive control designs.
>
> ...In fact, the mere size of the BF for the passive control studies should be enough to warrant a critical eye to those studies, especially given the a priori uncertainty surrounding the question of whether WM training can improve Gf. This leaves us with the 12 active control studies, for which (a) the Bayes factors for the individual studies overwhelmingly favor the null, (b) the meta-analytic BF favors the null, (c) the estimated effect sizes are not different from zero, and (d) half of the studies show raw effect sizes indicating a negative effect of transfer.
>
> ...The hierarchical Bayesian models suggest a two-factor model for explaining training effects: One factor is the type of experimental design used by the researcher (active vs. passive control) and the other is country of origin of the study (USA vs. non-USA). We submit that the discrepancy between the active and passive controls is consistent with a placebo effect, and we suspect that the effect of country of origin reflects idiosyncratic differences in experimental methods between the USA and non-USA studies. Setting aside specific causal mechanisms for the observed pattern of effect sizes, it is clear that the data reflect two separate data-generating processes, neither of which can be attributed to n-back training...It should be noted, however, that choice of experimental design also covaried with whether the study was conducted within the USA or outside the USA. Most of the studies using active controls were conducted within the USA, whereas the majority of the studies conducted outside of the USA used passive controls. While this leaves open the possibility that cultural differences are driving the difference between the active and passive studies, we doubt cultural differences would account for the 7-fold increase in the training effect, especially since the non-USA studies were primarily conducted in Westernized cultures (e.g., Europe).
Much as expected.
> The authors thank Jacky Au and Susan Jaeggi for sharing their data and for providing details of their analysis.
Almost a year later, Au still hasn't sent me the data, incidentally.
--
gwern
http://www.gwern.net
/ http://sci-hub.org/downloads/1b56/dougherty2015.pdf
> While we commend Au et al. on a rigorous meta-analysis, we contend that their analysis insufficiently address these issues. For example, while Au et al. relied on well-established null hypothesis significance testing (NHST) methods for meta-analysis, two well-known limitations of the NHST framework are that it tends to overstate evidence for the alternative hypothesis and does not permit one to evaluate the relative probability that the null hypothesis is in fact true. 1 In the context of the WM training literature, both of these problems are especially salient because the primary issue of debate is if working memory training is effective at all. This implies a need to evaluate the degree to which the data support the alternative hypothesis relative to the null, and is most easily addressed within a Bayesian approach.
Especially if you use an informative prior for how often intelligence
interventions fail to boost scores on IQ tests, and how often IQ test
score increases then fail to turn out to be on the latent g factor.
> Au et al. (2015) made an excellent attempt to reduce the potential influence of publication bias, with many studies included from nonpublished reports. The selection of studies to be included in the analysis appears to have been thorough and fair.
lol no.
> Au et al. (2015) presented effect sizes for 24 individual comparisons drawn from 20 papers. The aggregate weighted effect size across these 24 comparisons was 0.24. They also evaluated several possible mediators, including whether the studies used an active control (N = 12) or a passive control (N = 12), which yielded effect sizes of 0.06 and 0.44, respectively. Although Au et al. reported this effect as significant, they concluded that type of control group did not moderate the effect. This strikes us as an odd conclusion given that the magnitudes of these effect sizes differ considerably. Au et al.’s conclusion was based on a comparison between the control groups for active and passive studies, not by comparing the control groups to the treatment condition. The comparison of control groups while ignoring the training groups isn’t particularly informative regarding effect of training, since the effects of training can only be assessed relative to the control. In this regard, it is interesting to note that the effect size for the training condition amongst active-control studies (d = 0.25) is actually numerically smaller than the effect size amongst the control participants in the passive control studies (d = 0.28). The question is: Do these effect sizes provide evidence for training effectiveness?
Indeed. But in Au's defense, they claimed that an interaction with
non-American samples is driving the inflation of passive control
groups, not that the inflation exists (as it obviously does just
looking at the chart).
> The first step of our analysis involves transforming the effect sizes presented in Fig. 3 of Au et al. to their corresponding t values using t = sqrt(1/n1 + 1/n2) * g, where g is the measure of effect size and n1 and n2 are the sample sizes for two independent groups used in the effect-size calculations. We then computed the default Bayes factor (BF) corresponding to each t statistic using the ttestBF function in the BayesFactor package in R (Morey, Rouder, & Jamil, 2014; R Core Team, 2014) as well as the meta-analytic Bayes factor using the meta.ttestBF function. For all analyses, we set the scale factor on effect size to r = 1 and used a one-sided interval, which places the mass of the prior on effects greater than zero. The one-sided test is a reasonable assumption under the hypothesis that training should lead to improvements in Gf. Importantly, even large modifications to the prior distribution do not alter our conclusions in any substantive way, nor does using a two-sided null interval.
>
> ...As should be evident from Fig. 1 and Table 1, few of the individual studies provide particularly strong evidence for either the null or the alternative. Yet, looking across the entirety of the results, a curious pattern is obvious. First, 11 of the 12 effect sizes for the passive control studies are positive, whereas only 6 of the 12 effect sizes are positive for the active-control studies. Second, when these effect sizes are evaluated in terms of the Bayes factor, the majority of the individual studies favor the null hypothesis, including 6 of the 12 passive-control studies. These individual results using the BF roughly mirror the conclusions drawn from the significance tests, though the BF illustrates that the bulk of the studies show evidence for the null. However, these individual comparisons do not capitalize on a major strength of meta-analytic techniques, which is the ability to aggregate across studies to overcome the sample size problem. Moving on to the meta-analytic results, here the results diverge somewhat from the conclusions garnered from the individual studies. First, ignoring the type of control, the odds in favor of the alternative hypothesis is 152:1. This qualifies as 'decisive' evidence according to Jeffreys’ (1961) scheme. Figure 2, which provides the BFs conditioned on the use of passive- versus active-control groups, paints a much different picture. While the Bayes factor for the passive control studies is a whopping 13,241:1 in favor of the alternative, the Bayes factor for the active control studies is a more modest 7.7:1 in favor of the null.
A hypothesis-testing framework is ugly though, and makes it harder to
examine Au et al's international claim, so:
> Thus, we conducted a series of follow-up analyses using hierarchical Bayesian modeling, in which we modeled the effect sizes as a function of control group type (passive vs. active) as well as an additive effect of both control group type and country of origin (USA vs. non- USA). Au et al. (2015) identified country of origin as an im- portant moderator variable, with studies conducted within the USA yielding a small nonsignificant effect size and studies conducted outside the USA resulting in a moderate significant effect size – an effect that Au et al. hypothesized could be due to differences in motivation or compliance between USA and non-USA subjects. The inclusion of country of origin in our analysis allowed us to control for a potential important source of variability that Au et al. (2015) felt was theoretically justified. As we illustrate, inclusion of this variable in the Bayesian model reveals that the only estimated effect sizes that are different from zero are those based on non-USA passive-control studies. Furthermore, the estimated effect size for the active-control studies within the USA shrink to essentially zero...Strikingly, even when the prior distribution is set such that the effect of training is assumed to be large, there is still no evidence of that n-back training leads to improvements on Gf measures. While this model estimates that the median effect size amongst the active control studies is slightly above zero, this small positive effect is essentially eliminated when country of origin is added as a predictor in the model, as shown in Figs. 5 and 6. Importantly, the three international studies using active controls fail to yield a reliable positive effect. Furthermore, at the aggregate level the only effect size in which the HDI does not include zero are effects based on studies conducted outside the USA that use passive control designs.
>
> ...In fact, the mere size of the BF for the passive control studies should be enough to warrant a critical eye to those studies, especially given the a priori uncertainty surrounding the question of whether WM training can improve Gf. This leaves us with the 12 active control studies, for which (a) the Bayes factors for the individual studies overwhelmingly favor the null, (b) the meta-analytic BF favors the null, (c) the estimated effect sizes are not different from zero, and (d) half of the studies show raw effect sizes indicating a negative effect of transfer.
>
> ...The hierarchical Bayesian models suggest a two-factor model for explaining training effects: One factor is the type of experimental design used by the researcher (active vs. passive control) and the other is country of origin of the study (USA vs. non-USA). We submit that the discrepancy between the active and passive controls is consistent with a placebo effect, and we suspect that the effect of country of origin reflects idiosyncratic differences in experimental methods between the USA and non-USA studies. Setting aside specific causal mechanisms for the observed pattern of effect sizes, it is clear that the data reflect two separate data-generating processes, neither of which can be attributed to n-back training...It should be noted, however, that choice of experimental design also covaried with whether the study was conducted within the USA or outside the USA. Most of the studies using active controls were conducted within the USA, whereas the majority of the studies conducted outside of the USA used passive controls. While this leaves open the possibility that cultural differences are driving the difference between the active and passive studies, we doubt cultural differences would account for the 7-fold increase in the training effect, especially since the non-USA studies were primarily conducted in Westernized cultures (e.g., Europe).
Much as expected.
> The authors thank Jacky Au and Susan Jaeggi for sharing their data and for providing details of their analysis.
Almost a year later, Au still hasn't sent me the data, incidentally.
--
gwern
http://www.gwern.net
Pontus Granström
Jun 19, 2015, 1:29:39 PM6/19/15
to Dual N-Back, Brain Training & Intelligence
So n-back has a placebo effect, big surprise.
--
You received this message because you are subscribed to the Google Groups "Dual N-Back, Brain Training & Intelligence" group.
To unsubscribe from this group and stop receiving emails from it, send an email to brain-trainin...@googlegroups.com.
To post to this group, send email to brain-t...@googlegroups.com.
Visit this group at http://groups.google.com/group/brain-training.
For more options, visit https://groups.google.com/d/optout.
Gwern Branwen
Jun 19, 2015, 3:21:01 PM6/19/15
to N-back
On Fri, Jun 19, 2015 at 1:29 PM, Pontus Granström <lepo...@gmail.com> wrote:
> So n-back has a placebo effect, big surprise.
I feel like you're not getting it. If n-back has a placebo effect
> So n-back has a placebo effect, big surprise.
which is simply people trying harder on the retest and without this
there is not even an IQ test score increase but simply increasing
evidence for no effects, then *n-back does nothing of value for
intelligence*.
And it's not like there's great evidence at the moment for WM training
in general transferring to non-WM tasks.
--
gwern
http://www.gwern.net
Mercel
Jun 19, 2015, 3:40:23 PM6/19/15
to brain-t...@googlegroups.com, gw...@gwern.net
Nothing better than support for the null!
jttoto2
Jun 19, 2015, 6:11:48 PM6/19/15
to brain-t...@googlegroups.com
Anyone still up for restructuring this mailing list to encompass other possible ways of improving cognition? It seems that n-back is a dead end, confirmed at least a year ago.
Message has been deleted
jttoto2
Jun 19, 2015, 6:43:54 PM6/19/15
to brain-t...@googlegroups.com
In any case, is there a decent explanation as to why there are greater effects to passive control groups vs. active control groups in regards to Dnb? This isn't rhetorical, I'm wondering if someone can provide a link. I'm curious to see if this is due to perceived training causing a greater placebo effect, or if this is a case where the placebo is not really a placebo.
Gwern Branwen
Jun 19, 2015, 6:50:12 PM6/19/15
to N-back
On Fri, Jun 19, 2015 at 6:29 PM, jttoto2 <john....@gmail.com> wrote:
> In any case, is there a decent explanation as to why there are greater
> effects to active control groups vs. passive control groups in regards to
> In any case, is there a decent explanation as to why there are greater
> Dnb? This isn't rhetorical, I'm wondering if someone can provide a link.
Motivational. People who are in the experimental group feel an
expectation to perform better, either because they buy into brain
training or they don't want to let the nice researchers down and ruin
their experiment (don't knock this as an effect; I've participated in
several studies and have felt exactly this), and vice versa, the
people in the passive control group know nothing is expected of them
so why try hard?
The observed effect of half an SD is within the range demonstrated as
possible by past research and implied by the g-loading not being
unity; we've discussed in the past an experiment by Duckworth showing
that monetary incentives could boost IQ test scores by somewhere
around half an SD, IIRC.
--
gwern
http://www.gwern.net
☉
Jun 19, 2015, 8:41:45 PM6/19/15
to brain-t...@googlegroups.com, gw...@gwern.net
Generally speaking, a study should be designed to eliminate this problem. Why tell people which group they're in at all?
argumzio
☉
Jun 19, 2015, 9:07:06 PM6/19/15
to brain-t...@googlegroups.com
That would be off-topic for this thread, but besides improving general health and tDCS (and similar technologies), I don't think there's anything out there generally shown to achieve the Limitless result. At this stage, all that really remains is learning additional skills, like programming - and that amounts to interfacing with computer hardware. Ultimately, integrating with technology is the only other avenue, which branches off into many possibilities, the most mundane example of which would be reading and writing, the pinnacle of which would be programming, I think.
The limiting factor to learning such skills, then, becomes how long one can stick around in this mortal realm...
But what does one mean by "improving cognition"? What cognition would have to be improved?
argumzio
Pontus Granström
Jun 20, 2015, 3:59:34 AM6/20/15
to Dual N-Back, Brain Training & Intelligence
Is it important to you Gwern that only the genetic component changes to make it a "real" gain in IQ? Thickness of the brain or whatever, I dont understand your reasoning?!
Gwern Branwen
Jun 20, 2015, 11:59:00 AM6/20/15
to N-back
On Sat, Jun 20, 2015 at 3:59 AM, Pontus Granström <lepo...@gmail.com> wrote:
> Is it important to you Gwern that only the genetic component changes to make
> it a "real" gain in IQ?
When I write 'latent g factor', it does not mean genetics, and I'm
> Is it important to you Gwern that only the genetic component changes to make
> it a "real" gain in IQ?
getting a little tired of how you persist in denying all the null
results and criticisms while not bothering to understand the concepts
despite my years of providing fulltext of the relevant papers and even
excerpting the key passages to make them as easy to understand as
possible.
--
gwern
http://www.gwern.net
αrgvmziΩ
Jun 20, 2015, 1:28:23 PM6/20/15
to brain-t...@googlegroups.com
Do you think, gwern, that it would be more charitable to say that Pontus cannot understand, instead of supposing he persists in denial? Which is more likely: malice or stupidity?
argumzio
Brandon Woodson
Jun 20, 2015, 1:48:36 PM6/20/15
to brain-t...@googlegroups.com
At the risk of seeming as though I'm teaming against Pontus (trust me, I'm not; anyone is entitled to believe, at the peril of self and others, what they wish, whether it's true or not, unfortunately), I'm almost compelled to ask what the difference is? :)
--
Pontus Granström
Jun 20, 2015, 2:23:54 PM6/20/15
to Dual N-Back, Brain Training & Intelligence
I dont really have the energy to argue with you aspberger people.....
Brandon Woodson
Jun 20, 2015, 2:35:13 PM6/20/15
to brain-t...@googlegroups.com
I really wasn't being facetious in the preface to my question (that was in fact supposed to be a statement)... I hope it wasn't taken that way. Regards.
Gwern Branwen
Jun 20, 2015, 2:55:45 PM6/20/15
to N-back
On Sat, Jun 20, 2015 at 2:23 PM, Pontus Granström <lepo...@gmail.com> wrote:
> I dont really have the energy to argue with you aspberger people.....
You don't have the energy to understand the basic concepts of
> I dont really have the energy to argue with you aspberger people.....
something you spent a PhD thesis and years working on demonstrating?
You personally recruited dozens of people and wasted their time and
your own and money trying to show transfer from n-back training to
intelligence and you cannot be bothered to understand what
intelligence is or how one would show gains or what methodological
traps there are? Have you no shame?
I suppose I should be grateful that I've been upgraded from fascist
psychopath to 'aspberger', though.
--
gwern
http://www.gwern.net
Pontus Granström
Jun 20, 2015, 3:04:51 PM6/20/15
to Dual N-Back, Brain Training & Intelligence
Psychopaths and aspberger share much variance. They both have a hard time understanding the feelings of others, got rigid views etc. Very much the mind of you and your little friend Argumzio. You have this superior attitude (you know everything best and everyone one else is wrong, science is wrong etc, almost laughable), and you got your "uberfaq" to support your "uber-thinking". There are far worse things in this world than n-back training. If you want to save the world do something else than create this uberfaq targetting n-back. It's almost laughable that n-back research is a lack of shame, when I can name far worse things that people spend money on. Junk food, cigarettes, alchohol....
Gwern Branwen
Jun 20, 2015, 3:15:26 PM6/20/15
to N-back
On Sat, Jun 20, 2015 at 3:04 PM, Pontus Granström <lepo...@gmail.com> wrote:
> Psychopaths and aspberger share much variance.
No, they don't. They don't correlate, and they don't share any
> Psychopaths and aspberger share much variance.
mechanisms as far as we know.
> They both have a hard time
> understanding the feelings of others, got rigid views etc.
understanding and manipulation is one of the central traits
highlighted by all observers going back to Cleckley.
> There are far worse things in this world than n-back
> training. If you want to save the world do something else than create this
> uberfaq targetting n-back. It's almost laughable that n-back research is a
> lack of shame, when I can name far worse things that people spend money on.
> Junk food, cigarettes, alchohol....
to the worst of arguments: never mind that you had an entire field to
choose from and you could have done real research of value, could have
consumed those resources for something of nonzero value, you will
instead defend yourself, like a child, by pointing out that other
people have done worse. ('Mommy, but Jack down the street gets to stay
up late and eat junk food! why do I have to eat my carrots and go to
bed on time?')
But those people who eat junk food do not pretend it is virtuous, do
not publish about their junk food, do not receive public subsidies and
funding for their junk food or work at universities about their junk
food, do not waste the time of public-spirited volunteers by feeding
them junk food and claiming it may increase their intelligence, and so
on. Even when you acknowledge the failure and waste, you refuse to
learn anything from it and defend the defenseless; what can we call
this but shamelessness?
The only thing of value you have done is furnish a cautionary example
about the consequences of contempt for history, for statistics, for
methodology, and a lesson in what science is not.
--
gwern
http://www.gwern.net
Zaraki
Jun 20, 2015, 3:53:38 PM6/20/15
to brain-t...@googlegroups.com
Den lördag 20 juni 2015 kl. 20:23:54 UTC+2 skrev King Of The Stars:
I dont really have the energy to argue with you aspberger people.....
Please, it is spelled 'asperger', and I suggest you don't use it as a derrogatory term unless you want to offend alot more people than Gwern and Argumzio.
Pontus Granström
Jun 20, 2015, 4:01:31 PM6/20/15
to Dual N-Back, Brain Training & Intelligence
I apologize but gwern and argumzio are just too much sometimes. As some one said before, they are politically motivated.
--
αrgvmziΩ
Jun 20, 2015, 4:16:52 PM6/20/15
to brain-t...@googlegroups.com
Considering this baseless verbal attack (what is my political motivation?) I think Pontus' malicious character is plenty obvious. It's rather amusing that when facts and evidence are provided and discussed, the only response Pontus Granström has consistently provided over the years ranges from utter irrelevance to illogical personal attacks.
It's OK Pontus, we know you believe you really improved (what?) thanks to n-back, but at least contemplate putting your money where your mouth is. If you were an example of any kind of effect of n-back, I submit it would best approximate misplaced aggression and vindictiveness.
And since I am not on firstname terms with gwern, we have never been friends.
Here's to hoping you man up and knock it off with your campaigns against forum members who better appreciate facts, analysis, and understanding instead of slovenly aggression and delusion.
argumzio
Mercel
Jun 20, 2015, 4:57:30 PM6/20/15
to brain-t...@googlegroups.com, gw...@gwern.net
There's a tremendous ambiguity with regards to the findings from the research done on psychopathy; some papers have found that psychopaths have decreased accuracy in the judging of facial expressions, similar to those suffering from AS, while others have not found this. In any case, your cocksureness is unwarranted.
Message has been deleted
jttoto2
Jun 20, 2015, 9:45:56 PM6/20/15
to brain-t...@googlegroups.com
I apologize but gwern and argumzio are just too much sometimes. As some one said before, they are politically motivated.
I don't want to put any more salt on the wound, but the word political
tends to be associated with matters concerning public policy. There is
a difference between putting stock in reason and data, and holding on to a belief despite contrary evidence. If you want to call the former rigid, then so be it. I will likely not
give any rebuttal to your response, mainly because this thread is
inducing a flashback about a certain 3 page long debate with you
(remember those times?), but can you just stop for one minute, and just
acknowledge that the data does not support DnB at the moment? This is
based on a several papers posted in this mailing list. I mean, what is
there to gain from this?
At the very least, it is something to consider.
At the very least, it is something to consider.
Bobe Dilyan
Jun 21, 2015, 5:11:57 AM6/21/15
to brain-t...@googlegroups.com
Gwern is unusually inflexible and bigoted. If this is due to his condition then it is relevant to any intellectual discussion he may participate in. For example, he can't comprehend video formats and lacks the capacity to understand relative values.
Since his disability impairs his judgement, he may wish to avoid discussing this subject further. I view this as no different from asking people profiting off of N-Back or the scientific racist community (one of whom shot up a South Carolina church a couple of days ago) to step aside from any discussions on cognitive enhancement. Also, he clearly is a very bitter person and oftentimes derives preposterous outcomes from scientific literature to satiate his wounded ego. Gwern's biased interpretation of data really threw me off my own ambitions to understand working memory and cognitive enhancement. But when I actually took the time and dug into some of his claims I found them wildly misleading.
A quick hits of just some of his deficiencies:
-Mischaracterized a scientific study on cognitive training for ALS patients.
-Is vocally bitter about his own "failed" experiences with Dual N-Back. Likely blames DNB and Jaeggi personally for failures in his personal life.
-Refused to watch a video by Jason Chein, "if he wants to defend Dual N-Back, he can do it in print." then says that his disability prevents him from understanding videos.
-Is a proponent of eugenics and likely a racialist. Known to commune in HBD hangouts.
-Has an absolutely stupid argument against continuing research into working memory training for cognitive enhancement. The relative importance of cognitive enhancement is such that any significant or positive finding means that work should be pursued in the area. Cognitive enhancement is like cancer research, it is absolutely essential to human prosperity and therefore deserving of a greater share of time and resources. HIs perspective is that since there was only a minor positive correlation between working memory training and fluid G using dual n back for a short period of time using "active" controls that everybody should pack up their bags and go home. Perhaps its his aspergers disability, but this in no way should engender hostility to pursuing working memory training as a means of cognitive enhancement. In fact, the research is bright enough to continue studying working memory over the long term and pursuing permutations of working memory training that could result in higher G transfer, which is exactly what someone like Jason Chein is doing.
αrgvmziΩ
Jun 21, 2015, 11:28:05 AM6/21/15
to brain-t...@googlegroups.com
0/10. Your complaint generator is showing, troll.
argumzio
--
Mercel
Jun 21, 2015, 12:31:01 PM6/21/15
to brain-t...@googlegroups.com
If we're honest about the fact that much of science is desire driven in the Nietzschean sense, then I have no problem with Gwern's pre-scientific orientation or proclivity towards certain values or conceptions as long as we always remain within the bounds of science. We are concerned with truth, but it is always desirable as the primary point of departure for scientific inquiry that this truth converge with our own conception of the real before we grant something to be real.
Message has been deleted
Message has been deleted
Message has been deleted
jttoto2
Jun 21, 2015, 1:28:37 PM6/21/15
to brain-t...@googlegroups.com
I
can't, for the life of me, understand how one can type about one
marginalized group (blacks), while simultaneously disparaging another
marginalized group (Those with Asperger's, and by extension those with
autism ) in the same post, all in the name of tabloid garbage.
Bobe Dilyan
Jun 21, 2015, 3:51:06 PM6/21/15
to brain-t...@googlegroups.com
An emotional response. I'm saying that his bias against blacks and his Asperger's likely color his reasoning. That's not tabloid.
@Mercel Interesting that you say Gwern is a pre-scientific Nietzchean. He reminds me of Nietzsche's Last Man desperately seeking to destroy that which is good because it disagrees with his internal temperature.
Mercel
Jun 21, 2015, 4:36:23 PM6/21/15
to brain-t...@googlegroups.com
Actually, I'm not saying that Gwern himself has anything to do with Nietzche, but that science itself is Nietzchean in the sense of will to power; before truth is truth as truth, we hope that the truth is a kind of truth that conforms to our pre-scientific world-conception. It's not that we are purely objective beings wanting to find some kind of objective truth (the classical, naive conception of the scientist who merely wants to find the truth -- any kind of truth out there), but beings as hermeneutically oriented beings, beings who want our own personal conception of reality made into objective truth. There's an important and rarely talked about gap between truth and subjective conception of truth within scientific enterprise; before science is science, it is in a sense ideology, philosophy or personal beliefs that steers science, science in its immature phase, in one direction rather than another.
I remember once when Gwern tried to seemingly guiltify me by associating me with Marxist terminology (the term "false consciousness") in some post not too long ago, and that's fundamentally not a problem -- I very much would have wanted to see the world as a one big possibility, that a person could mend his own intellectual wounds afflicted on him by a past which we had no control over and better his position and chances in this world by using programs such as n-back. I dislike determinism and love freedom and so on and so on. Cognitive enhancement is very much a grey area of science where non-scientific values and conceptions can play themselves out.
☉
Jun 21, 2015, 5:08:23 PM6/21/15
to brain-t...@googlegroups.com
There's a lot to be said for not feeding the trolls.
Sure, gwern has pissed off more than his fair share of people on the internet, and that might play itself out in time, but this is not the place to respond to trolls who use Scott Pakin's complaint generator as if it were genuinely meaningful.
Ersatz vilification and hate mongering do not deserve attention. And the fact that individuals engage in such behavior here should speak to the results associated with n-back research.
Here's to hoping a troll won't come around telling us there's no absolute truth and that relativism is the only real truth. What utter bullshit.
argumzio
Bobe Dilyan
Jun 21, 2015, 7:22:26 PM6/21/15
to brain-t...@googlegroups.com
What's interesting is that I have made very concrete criticisms towards Gwern's analytical powers and you chose to focus on the syntactic sugar the criticism was wrapped in. There is no doubt that Gwern believes in his righteous cause of retribution, and that he has followers thinking the same. But when will these followers step outside of their ideology and address the criticisms I have made? If Gwern is to be trusted as any sort of source, he needs to have the genitalia (or fortitude, if overt masculine language frightens you) to address his limitations plainly splattered across this mailing list.
Part of my problem with Gwern is that I trusted his meta-analysis. I assumed that the N-Back was a waste of time, a joke, or a scam. Only when I read into the research did I discover that it was in fact a promising beginning into the workings of human cognition and intellectual advancement. It was his personal incredulity seasoned with limited null findings that distracted me from reading deeper in the literature. Maybe it's my fault, or maybe I expected somebody so self assured to actually be right, but he has no resources to speak meaningfully on this topic. Therefore, for him to guide the direction of this mailing list is a complete farce.
@Mercel I think the politicization of this field is disappointing. On the one hand, you have the "self help" crowd believing in the power of anything to change everything. On the other hand, you have the effete privileged classes believing in their own unchangeable superiority. That being said, I think it's unavoidable. Each one of us approaches this topic with our subjective power interests in tow. This is why it's necessary to engage in conversations that some find distasteful to their precious world views.
☉
Jun 21, 2015, 9:07:23 PM6/21/15
to brain-t...@googlegroups.com
"Syntactic sugar" my ass. Where's the concrete criticism? All mere farcical bluster and brown noise.
So gwern has put a lot of effort into what he's done and you just can't handle it. Despite your apparent pretensions to knowledge of "the literature", verbal assaults, innuendo, and inane drivel as you have here spewed do not make a strong case in your favor. Quite the opposite.
Let's all happily remember those halcyon days in this forum when ... oh who am I kidding? It's always been the case here that people who believe they can become paragons of intellect through n-back and brain training were always those most vociferous in defending all the time and effort they spent doing those tasks. And who here hasn't spent time doing them?
1/10 now. Troll harder, Bobe Dilyan. I think that you might actually manage to fool a few people into believing that your dissatisfaction stems from reasoned analysis and thorough comprehension of the literature. Why would anyone go on in your manner unless you had the balls (or brains?) to back it up, right? A real alpha internet tough guy. Say enough shit, it might manage to stick. You might even manage to pick and choose your targets in this list over the coming months. I might even get enough spare time to dismantle a fair portion of whatever it is you post, assuming you post something that merits such a response as I have provided for others before.
Whatever the case, Bobe Dilyan, maybe you should just put your nose to that grind wheel with QNB for over an hour at a stretch for a month. Y'know, if you have the balls. ;)
argumzio,
why I check my privilege six times before breakfast
Bobe Dilyan
Jun 21, 2015, 11:24:26 PM6/21/15
to brain-t...@googlegroups.com
You didn't address a single criticism I broached earlier in the thread. Since that was directed towards Gwern, I'll assume you lack the ability to address those points and just want to yell about whatever you think I am saying.
Should a person do DNB? Exclusively, I'm not sure if short term DNB training results in major gains in pattern recognition. I think this meta-analysis suggests this. Some variation of DNB in combination with other cognitive performance boosters? Absolutely the best bet at this moment in time. Since we know that there is a positive correlation between memory training/video game skills and near to far transfer on a range of abilities, there must be some internal mechanism of the brain that responds to prompted stimuli which can be accessed for other tasks. I believe Chein's theory of strategy vs core training can address some of the discrepancy in outcomes between individuals. So his working memory training task would be the one I would select to incorporate into a cognitive enhancement regimen.
Most people that have cognitive deficiencies simply need to start exercising more and reading in the evening. That's the place to start. DNB and working memory training paradigms come later. Nootropics and TDCS would be the last step to recommend (even though I am very skeptical of both methods). This says nothing about other interventions/training paradigms of which I am not familiar.
☉
Jun 21, 2015, 11:57:20 PM6/21/15
to brain-t...@googlegroups.com
I'll conclude beyond the shadow of a doubt you're not worth my time.
argumzio
Bobe Dilyan
Jun 22, 2015, 12:24:17 AM6/22/15
to brain-t...@googlegroups.com
Fantastic. Feel free to exit this google group permanently.
☉
Jun 22, 2015, 12:30:42 AM6/22/15
to brain-t...@googlegroups.com
Feel free not to overstay your welcome. :)
argumzio
Hi
Jun 22, 2015, 1:21:18 AM6/22/15
to brain-t...@googlegroups.com
June 2015
Despite the popularity of working memory (WM) and updating training, recent reviews have questioned their efficacy. We evaluated a computer-based training programme based on the Running Span and Keep Track paradigms. We assigned 111 7-year-olds with poor WM and mathematical performances to updating training, one of the two control groups, or a fourth group, who were administered Cogmed, a commercially available programme. At the immediate posttest, updating training produced only marginal improvements relative to control, but this was sustained and became significant six months post-training. Cogmed training resulted in substantial improvement at immediate posttest, but became marginal at delayed posttest. Neither type of training resulted in better performance in mathematics or generalised to other WM tasks that differed more markedly from those used during training. These findings suggest that relations between WM or updating capacity and mathematics performance may be moderated by factors that do not benefit directly from improved capacity.
Hi
Jun 22, 2015, 1:23:10 AM6/22/15
to brain-t...@googlegroups.com
The neuroplastic effect of working memory training in healthy volunteers and patients with schizophrenia: Implications for cognitive rehabilitation
August 2015
Abstract
We conducted an activation likelihood estimation (ALE) meta-analysis to quantitatively review the existing working memory (WM) training studies that investigated neural activation changes both in healthy individuals and patients with schizophrenia. ALE analysis of studies in healthy individuals indicates a widespread distribution of activation changes with WM training in the frontal and parietal regions, especially the dorsolateral prefrontal cortex, the medial frontal cortex and the precuneus, as well as subcortical regions such as the insula and the striatum. WM training is also accompanied by activation changes in patients with schizophrenia, mainly in the dorsolateral prefrontal cortex, the precuneus and the fusiform gyrus. Our results demonstrate that WM training is accompanied by changes in neural activation patterns in healthy individuals, which may provide the basis for understanding neuroplastic changes in patients with schizophrenia.
On Monday, June 22, 2015 at 2:30:42 PM UTC+10, ☉ wrote:
αrgvmziΩ
Jun 22, 2015, 2:03:19 AM6/22/15
to brain-t...@googlegroups.com
If a brain did not change in response to the environment, it wouldn't be doing its job now, would it? Neuroplasticity can only go so far when there are constraints associated with the optimization taking place neurologically.
At this point, it would seem the abstract is basically describing the neurocorrelates of crystalization (skill acquisition) for the task. It is still very doubtful that there is any meaningful remedial effect as far as gF is concerned, assuming such is feasible at all. If the body of the article suggests differently, feel free to alert me to the relevant segments in the article.
And do tell us if you understand the article differently, should you feel so inclined...
argumzio
why sometimes the human can be all too human six times before breakfast
Hi
Jun 22, 2015, 2:35:17 AM6/22/15
to brain-t...@googlegroups.com
Oh I didn't post the articles to infer that I was on one side of the camp versus another, more that I'm not convinced that its black and white yet. I do think that an extremely strong case can be made for the so far lack of success to create a complex enough task to test such related hypotheses however.
Thanks for that though argumzio, appreciated.
To me, although I could be wrong, all games I've encountered so far that aim to increase cognition are analogous to researchers training people to do push ups in order to prepare for a body building competition, therefore any evidence relating to a for or against claim to me is treated with a grain of salt. However, given any positive results, it only makes sense to me that my more advanced adaptations (for example, if used the pattern principle of dual-n-back and applied it to other mental techniques) would likely only be met with greater of the positive claim, relative to the tapped domains. Again, I could be wrong but that's to my logic (of course I realise things can be counter intuitive and that this could be one of them) so far. Regardless, all in all, because I'm applying it to life sciences/situations its obviously got a lot of benefit outside the parameters directly related to improving cognition specifically more than cognitive skills, especially when I'm incorporating it with "Memory Championship" like memory techniques.
Reply all
Reply to author
Forward
0 new messages