Wednesday, April 24, 2019

The Two Cultures of Computational Psychiatry ( tools from cognitive science, computational neuroscience, and machine learning)

The Two Cultures of Computational Psychiatry. Daniel Bennett, Steven M. Silverstein, Yael Niv. JAMA Psychiatry. Published online April 24, 2019, doi:10.1001/jamapsychiatry.2019.0231

Computational psychiatry is a rapidly growing field that uses tools from cognitive science, computational neuroscience, and machine learning to address difficult psychiatric questions. Its great promise is that these tools will improve psychiatric diagnosis and treatment while also helping to explain the causes of psychiatric illness.1-3

Within computational psychiatry, there are distinct research cultures with distinct computational tools and research goals: machine learning and explanatory modeling.1 While each can potentially advance psychiatric research, important distinctions between the cultures sometimes go unappreciated in the broader psychiatric research community. We detail these distinctions, referring to Breiman’s influential dichotomy between these cultures of statistical modeling4 to identify limitations on the inferences that each culture can draw.

Breiman4 defined the 2 cultures of statistical modeling in terms of a data-generating process that generates output data from input variables. His dichotomy distinguished “algorithmic modeling,”4(p200) which aims to predict what outputs a data-generating process will produce from a given set of inputs while treating the process itself as a black box,2,3 from “data modeling,”4(p199) which uses the pattern of outputs and inputs to explain how the data-generating process works. In psychiatry, the data-generating processes are the psychological and neurobiological mechanisms that produce psychiatric illnesses. The output data produced by these processes are psychiatric outcomes (eg, symptoms, medication response) with input variables including family history, precipitating life events, and others. Breiman’s distinction between prediction and explanation is also what separates machine-learning approaches to computational psychiatry, which aim to predict psychiatric outcomes, from explanatory modeling, which aims to explain the computational-biological mechanisms of psychiatric illnesses. [...].

The culture of machine learning typically uses statistical techniques, such as support vector machines or deep neural networks, to predict psychiatric outcomes. These tools can be seen as lying on a continuum with classical statistics such as regression but with the addition of practices designed to reduce overfitting, such as parameter regularization and cross validation. For instance, a study by Webb et al5 has used such tools to predict antidepressant response from a combination of variables, including demographic factors, symptom severity, and cognitive task performance. Despite good predictive performance, the study drew no conclusions about the mechanisms by which these variables were linked to antidepressant response. This is because in machine learning, the parameters of the models that are used to predict psychiatric outcomes are not assumed to correspond to any underlying psychological or neural process; consequently, these parameters cannot be interpreted mechanistically. [...].

In comparison, the culture of explanatory modeling focuses on statistical models (expressed as equations) that define interacting processes with parameters that putatively correspond to neural computations. For instance, equations describing value updating in reinforcement-learning models are thought to correspond to corticostriatal synaptic modifications modulated by dopaminergic signaling of reward prediction errors. Consequently, explanatory model parameters fit to behavioral and/or neural data from patients with psychiatric diagnoses can directly inform inferences about dysfunctions in underlying neural computations, subject to several conditions being met. For instance, Huys et al6 have shown that anhedonia is correlated across diagnoses with a model parameter corresponding to the blunting of experienced reward value but not with a parameter controlling the rate of learning from this experienced value, providing evidence against one dopaminergic explanation of depression.

Importantly, there are several conditions that must be met before an explanatory model can be used in this way. First, to support the model’s correspondence to the true data-generating process and distinguish between different candidate models, the models must make sufficiently different predictions for the experimental data. Separately, to identify the model parameters accurately, the parameters’ effects on model predictions should be relatively independent, and there must be sufficient data. One approach to testing these conditions is to simulate data from each candidate model and test the ability of a model-fitting routine to recover the true cognitive model and its parameters from these data. Because empirical data will not correspond as perfectly to any of the candidate models, this test is a necessary but not sufficient condition for reliable explanatory modeling. Indeed, a common error is to overinterpret results [...].

A potential limitation of explanatory modeling in computational psychiatry is that theories (ie, models) may be ill-matched to available data, because data collected for other purposes may not distinguish between subtly (but importantly) different hypotheses regarding the mechanisms underlying psychiatric dysfunction. It is therefore crucial that explanatory modeling studies be carefully designed to ensure they provide data that can be used to accurately identify model parameters and discriminate between models. Another pitfall is the ubiquity of generalized performance deficits in individuals with mental illness, owing to factors such as low motivation, poor understanding of task instructions, and medication-induced sedation. [...].

Although these limitations mean that explanatory modeling can be a challenging enterprise, its potential benefits are also great. One exciting possibility is that parameters from explanatory models can be used as computational markers of psychiatric illness.1 Using such markers, it may be possible to (1) distinguish diagnoses that might initially have similar symptom profiles, such as major depression and bipolar disorder; (2) characterize within-diagnosis heterogeneity (and potentially generate new diagnostic categories) with reference to the disordered computational mechanism; or even (3) predict relapse and/or treatment responses based on shifts in computational markers. [...].

Conclusions

Applying Breiman’s dichotomy4 between the cultures of statistical modeling to computational psychiatry helps to parse the promises of this growing field. Crucially, it suggests that the 2 cultures of computational psychiatry are fundamentally suited for drawing different kinds of inferences from psychiatric data. This marks a point of difference between our dichotomy and Breiman’s dichotomy.4 Whereas Breiman espoused the virtues of prediction over explanation, we wish to emphasize the value of both cultures in asking distinct research questions and the importance of ongoing crosstalk between cultures. Although we have treated these cultures as separate, hybrid approaches1,2 have already proven powerful: generative embedding approaches incorporate parameter estimates from explanatory models as variables in machine-learning algorithms,7 and clusters of symptoms identified using machine-learning approaches can prompt explanatory modeling to determine the mechanisms underlying each specific cluster.2 Indeed, as long as we remain far from understanding the provenance of mental illness, it behooves us to use all appropriate methods to their full extent.

Full article and references at the link above

Why we devote ourselves to knowledge, the arts and the sciences

1  Caspar Heineman Tumblr blog, Nov 4 2012, https://angstravaganza.tumblr.com/post/34991232708/i-wasnt-always-smart-i-was-actually-very-stupid

“I wasn’t always smart, I was actually very stupid in school… [T]here was a boy who was very attractive who was even stupider than I was. And in order to ingratiate myself with this boy who was very beautiful, I began to do his homework for him – and that’s how I became smart, I had to do all this work to just keep ahead of him a little bit, in order to help him. In a sense, all the rest of my life I’ve been trying to do intellectual things that would attract beautiful boys.”

— Michel Foucault, 1983


https://quotefancy.com/quote/1492615/Leopold-von-Sacher-Masoch-Why-become-well-versed-in-science-and-the-arts-if-not-to

“Why become well-versed in science and the arts if not to impress a lovely little woman?”

— Leopold von Sacher-Masoch

Evolutionary Approaches to Complex, Asymmetrically Structured Societies; extending the gene/culture theory with a 3rd level, encompassing polities divided by class & ruled by elites

An evolutionary approach to complex hierarchical societies. Theodore Koditschek. Behavioural Processes, Volume 161, April 2019, Pages 117-128. https://doi.org/10.1016/j.beproc.2018.04.020

Highlights
•    Proposes A New Framework for Framing Evolutionary Approaches to Complex, Asymmetrically Structured Societies.
•    Draws on Gene/Culture ‘Dual Inheritance Co-evolutionary Theories from Evolutionary Anthropology.
•    Extends This Bi-Level Theory to Incorporate a Third Level, Encompassing Polities Divided by Class and Ruled by Elites.
•    Further Extends This Tri-Level Theory to Incorporate a Fourth Level Encompassing Modern, Economically Expansive Capitalist Societies.

Abstract: Evolutionary anthropologists have been remarkably successful in developing ‘dual inheritance’ theories of gene/culture coevolution that analyze the interaction of each of these factors without reducing either one to the other’s terms. However, efforts to extend this type of analysis to encompass complex, class-divided hierarchical societies, grounded in formal laws, political institutions, and trajectories of sustained economic development have scarcely begun. This article proposes a provisional framework for advancing such a multi-level co-evolutionary analysis that can encompass multiple forms of social organization from simple hunting/foraging groups to agrarian states and empires, up through the global capitalist system of our own day. The article formulates tools to conceptualize some of the ways in which ‘selection’ and ‘adaptation’ operate at every level to bring genes, cultures, states, and market exchange into provisional alignment with one another. It considers some of the ways in which modes of production’, ‘modes of coercion’ and ‘modes of persuasion’ interact complexly, at different societal levels.

---
How far can human nature, history, and society be explained in evolutionary terms? The question has long been controversial and has generated some of the most heated debates in the history of the social sciences. During the nineteenth century, many grandiose, sweeping claims were made in the affirmative, only to see the influence of such views decline in subsequent decades. Criticized for the racist, sexist, and elitist assumptions in which they were grounded, such evolutionary theorizing was further undermined by the modern synthesis in twentieth century biology, which focused on the role of genetics in shaping physical characteristics, implicitly leaving the study of human social behavior, social development, and organization to the non-evolutionary social sciences. During the 1960s and 70s, the rise of sociobiology threatened to up-end this consensus by proposing genetic explanations for behavior, but the biological reductionism of its leading promoters was vigorously resisted, and the traditional division of labor between the sciences of nature and of nurture continued to prevail. In evolutionary psychology and in human behavioral ecology, Darwinian processes are widely invoked, but in the remainder of the social sciences little attention to them is paid (Kevles, 1985; Degler, 1991; Segerstråle, 2000).

A major exception to this generalization can be found among a group of self-styled evolutionary anthropologists, who have devised a hybrid paradigm for bio-social theory that promises a completely novel way of restoring the evolutionary approach to human affairs. Eschewing any unilateral recourse to biological determinism, these evolutionary anthropologists have focused on the ways in which culture and biology interact. While certainly recognizing that human behavior is rooted in our heredity, they are concerned to understand how culture has co-evolved with our innate psychology, and how the two have mutually constructed one another in a manner that is unique among life forms. As they have pioneered this new approach to the nature/nurture conundrum, the evolutionary anthropologists have givenus the opportunity to break through a major impasse that has habitually tripped social scientists up. When we begin to apply their methods to the study of complex, asymmetrically structured societies, however, (such as those in which most humans through recorded history have lived) certain key modifications will need to be made. In this paper, I briefly outline these modifications and explore some of the ways in which they may be able to fortify the foundations on which a comprehensive evolutionary social science could be built.

---
Ugh:
Yet, none of these solutions proved to be permanent, as the conditions of competition and wealth creation changed. In the post-1975 period, full-scale globalization has unleashed forces that are now generating both massive economic growth, and also new types of destabilizing crisis. World trade and real Gross World Product have tripled during this period, approximately doubling real global per capita incomes during a periodwhen the rate of population increase has been slowing down. At the same time, these gains have been distributed very unevenly, as wages of less skilled workers in the developed countries have stagnated, pools of deep poverty and underdevelopment continue to fester (especially below the equator), and even in the richest countries, some 20% of the national income, and 35% of the national wealth,is monopolized by the top one percent. On every continent, irresistible surges of de-regulation and regressive taxation have jeopardized the viability of the welfare state, even as vast waves of migration have roiled labor markets everywhere around the globe. The result has been to exacerbate previous inequalities, and to disempower organized labor, pitting workers against one another, and fuelling the cultural politics of conflicting religions, ethnicities, and identities. Both dramatic economic growth and spectacular social dislocation have, overthe past generation, become the orders of the day (Piketty, 2014; Stiglitz, 2017; Data from Wikipedia entries on ‘World Population’, ‘Gross World Product’, and ‘Wealth Inequality in the United States’).

Hypergamy (a husband's earning capacity systematically exceeds that of his wife) is an important feature of Norwegian mating patterns, being Norway one of the most equalized societies

Almås, Ingvild & Kotsadam, Andreas & Moen, Espen R & Røed, Knut, 2019. "The Economics of Hypergamy," CEPR Discussion Papers 13606, https://ideas.repec.org/p/cpr/ceprdp/13606.html

Abstract: Partner selection is a vital feature of human behavior with important consequences for individuals, families, and society. Hypergamy occurs when a husband's earning capacity systematically exceeds that of his wife. We provide a theoretical framework that rationalizes hypergamy even in the absence of gender differences in the distribution of earnings capacity. Using parental earnings rank, a predetermined measure of earnings capacity that solves the simultaneity problem of matching affecting earnings outcomes, we show that hypergamy is an important feature of Norwegian mating patterns. A vignette experiment identifies gender differences in preferences that can explain the observed patterns.

Keywords:marriage, gender identity, labor supply, household specialization

Most organic evolution involves non-Darwinian processes; much behavioral selection is non-teleonomic; non-Darwinian processes provide a better model for most cultural evolution than does selection by consequences

The non-Darwinian evolution of behavers and behaviors. Peter R. Killeen. Behavioural Processes, Volume 161, April 2019, Pages 45-53. https://doi.org/10.1016/j.beproc.2017.12.024

Highlights
•    Most organic evolution involves non-Darwinian processes.
•    Much behavioral selection is non-teleonomic.
•    Non-Darwinian processes provide a better model for most cultural evolution than does selection by consequences.

Abstract: Many readers of this journal have been schooled in both Darwinian evolution and Skinnerian psychology, which have in common the vision of powerful control of their subjects by their sequalae. Individuals of species that generate more successful offspring come to dominate their habitat; responses of those individuals that generate more reinforcers come to dominate the repertoire of the individual in that context. This is unarguable. What is questionable is how large a role these forces of selection play in the larger landscape of existing organisms and the repertoires of their individuals. Here it is argued that non-Darwinian and non-Skinnerian selection play much larger roles in both than the reader may appreciate. The argument is based on the history of, and recent advances in, microbiology. Lessons from that history re-illuminate the three putative domains of selection by consequences: The evolution of species, response repertoires, and cultures. It is argued that before, beneath, and after the cosmically brief but crucial epoch of Darwinian evolution that shaped creatures such as ourselves, non-Darwinian forces pervade all three domains.

Keywords: Non-Darwinian evolution Selection by consequences Tree of life Lateral gene transfer Meme Microbes

What Might Have Been: Near Miss Experiences and Adjustment to a Terrorist Attack—narrowly avoiding a traumatic event leads to “survivor guilt” or distress over one’s comparative good fortune

What Might Have Been: Near Miss Experiences and Adjustment to a Terrorist Attack. Michael J. Poulin, Roxane Cohen Silver. Social Psychological and Personality Science, April 24, 2019. https://doi.org/10.1177/1948550619829064

Abstract: Near miss experiences—narrowly avoiding a traumatic event—are associated with distress, despite signaling good fortune. For some, near miss experiences call to mind those who, unlike oneself, were directly affected by the event, leading to “survivor guilt” or distress over one’s comparative good fortune. Survivor guilt, in turn, may function as upward counterfactual thinking about others’ negative outcomes, leading to intrusive thoughts and post-traumatic stress. We compared individuals who did or did not report a near miss with respect to the September 11, 2001, terrorist attacks—that is, almost being directly affected—in a national longitudinal study (N = 1,433). Near miss experiences predicted higher levels of reexperiencing symptoms and probable post-traumatic stress disorder, as well as maintenance of reexperiencing symptoms over the next 3 years. These associations were partially accounted for by survivor guilt. Near misses may be associated with distress in part because they entail reflection on negative outcomes for others.

Keywords: stress, trauma, meaning, terrorism

Humans and their working dogs: Comparing the human-dog socially distributed cognitive system with humans using non-biological tools and human interaction with draft animals

Distributed cognition criteria: Defined, operationalized, and applied to human-dog systems. Mary Jean Amon, Luis H. Favela. Behavioural Processes, Volume 162, May 2019, Pages 167-176. https://doi.org/10.1016/j.beproc.2019.03.001

Highlights
•    Distributed cognitive systems exhibit three criteria: interaction-dominant dynamics, agency, and shared task-orientation.
•    The three criteria allow for socially distributed cognitive systems to be distinguished from other group tasks.
•    Some interactions between domesticated dogs and humans can be properly characterized as distributed cognition.
•    Human-dog distributed cognitive systems are contrasted with extended cognitive systems (human-tool use) and obedience tasks.
•    These criteria and examples demonstrate the expansive scope of distributed cognition theory beyond verbal interactions.

Abstract: Distributed cognition generally refers to situations in which task requirements are shared among multiple agents or, potentially, off-loaded onto the environment. With few exceptions, socially distributed cognition has largely been discussed in terms of intraspecific interactions. This conception fails to capture some forms of group-level cognition among human and non-human animals that are not readily measured or explained in mentalistic or verbal terms. In response to these limitations, we argue for a more stringent set of empirically-verifiable criteria for assessing whether a system is an instance of distributed cognition: interaction-dominant dynamics, agency, and shared task orientation. We apply this framework to humans and working dogs, and contrast the human-dog socially distributed cognitive system with humans using non-biological tools and human interaction with draft animals. The human-dog system illustrates three operationalizable factors for classifying phenomena as socially distributed cognition and extends the framework to interspecies distributed cognition.

Keywords: Agency Distributed cognition Dogs Extended cognition Interaction dominance Working animal

Sculpting sex differences: Microglia are more phagocytic in the male amygdala during neonatal development; androgen-induced endocannabinoids increase phagocytosis in males, producing a sex difference in juvenile social play

Microglial Phagocytosis of Newborn Cells Is Induced by Endocannabinoids and Sculpts Sex Differences in Juvenile Rat Social Play. Jonathan W. VanRyzin et al. Neuron, Volume 102, Issue 2, 17 April 2019, Pages 435-449.e6. https://doi.org/10.1016/j.neuron.2019.02.006

Highlights
•    Microglia are more phagocytic in the male amygdala during neonatal development
•    Androgen-induced endocannabinoids increase phagocytosis in males
•    Microglia engulf viable newborn astrocytes in a complement-dependent manner
•    Developmental phagocytosis produces a sex difference in juvenile social play

Summary: Brain sex differences are established developmentally and generate enduring changes in circuitry and behavior. Steroid-mediated masculinization of the rat amygdala during perinatal development produces higher levels of juvenile rough-and-tumble play by males. This sex difference in social play is highly conserved across mammals, yet the mechanisms by which it is established are unknown. Here, we report that androgen-induced increases in endocannabinoid tone promote microglia phagocytosis during a critical period of amygdala development. Phagocytic microglia engulf more viable newborn cells in males; in females, less phagocytosis allows more astrocytes to survive to the juvenile age. Blocking complement-dependent phagocytosis in males increases astrocyte survival and prevents masculinization of play. Moreover, increased astrocyte density in the juvenile amygdala reduces neuronal excitation during play. These findings highlight novel mechanisms of brain development whereby endocannabinoids induce microglia phagocytosis to regulate newborn astrocyte number and shape the sexual differentiation of social circuitry and behavior.

We can be very good sisters and brothers with unknown, not-related-to-us guys if our family co-members are away: Grooming behavior of female Japanese macaques (Macaca fuscata yakui)

Behavioral responses to changes in group size and composition: a case study on grooming behavior of female Japanese macaques (Macaca fuscata yakui). Yosuke Kurihara, Mari Nishikawa, Koji Mochida. Behavioural Processes, Volume 162, May 2019, Pages 142-146. https://doi.org/10.1016/j.beproc.2019.03.005

•    Changes from multi-female to one-female groups were observed in Japanese macaques.
•    Adult females increased their grooming efforts in one-female periods.
•    They continued kin-biased grooming in one-female periods.
•    Unrelated juveniles were alternative grooming partners for them in one-female periods.
•    Rare observations of social changes are useful for understanding animal societies.

Abstract: Primates flexibly change their grooming behavior depending on group size and composition to maintain social relationships among group members. However, how drastic social changes influence their grooming behavior remains unclear. We observed the grooming behavior of adult female Japanese macaques in two groups temporarily formed as one-female groups from multi-female groups and compared their behaviors between the multi-female and one-female periods. Adult females more frequently performed grooming with both their relatives and unrelated juveniles during the one-female period when other adult females were unavailable as alternatives to their absent familiar partners. The increased grooming time and diversity of grooming partners might alleviate the short-term stress caused by the loss of grooming partners and reduce social instability or mitigate the long-term stress due to disadvantages in intergroup conflicts. Our study provides rare evidence on the flexibility in grooming behavior of primates and encourages accumulating case reports for understanding behavioral responses of primates to drastic social changes.

In rats: Resurgence of a target behavior suppressed by a combination of punishment and alternative reinforcement

Resurgence of a target behavior suppressed by a combination of punishment and alternative reinforcement. Rusty W. Nall, Jillian M. Rung, Timothy A. Shahan. Behavioural Processes, Volume 162, May 2019, Pages 177-183. https://doi.org/10.1016/j.beproc.2019.03.004

•    Initially, rats performed a target behavior for access to food.
•    Next, target behavior was punished and alternative behavior was reinforced.
•    Finally, target and alternative behavior were extinguished and resurgence occurred.
•    No resurgence occurred following punishment without alternative reinforcement.
•    The present results suggest implications for theory and applied treatment.

Abstract: Differential-reinforcement-based treatments involving extinction of target problem behavior and reinforcement of an alternative behavior are highly effective. However, extinction of problem behavior is sometimes difficult or contraindicated in clinical settings. In such cases, punishment instead of extinction may be used in combination with alternative reinforcement. Although it is well documented that omitting alternative reinforcement can produce recurrence (i.e., resurgence) of behavior previously suppressed by extinction plus alternative reinforcement, it remains unclear if resurgence similarly occurs for behavior previously suppressed by punishment plus alternative reinforcement. The present experiment examined this question with rats. In Phase 1, a target behavior (lever pressing) was reinforced with food pellets. In Phase 2, the target behavior continued to be reinforced, but it also produced mild foot shock and an alternative behavior (nose poking) also produced food. Finally, all consequences were removed and resurgence of target behavior occurred. Resurgence did not occur for another group that similarly received punishment of target behavior in Phase 2 but not alternative reinforcement. These results indicate that resurgence was a product of the history of exposure to and then removal of alternative reinforcement and that the removal of punishment alone did not produce resurgence of target behavior.

Keywords: Punishment Rats Relapse Resurgence