Approval-directed agency and the decision theory of Newcomb-like problems

Decision theorists disagree about how instrumentally rational agents, i.e., agents trying to achieve some goal, should behave in so-called Newcomb-like problems, with the main contenders being causal and evidential decision theory. Since the main goal of artificial intelligence research is to create machines that make instrumentally rational decisions, the disagreement pertains to this field. In addition to the more philosophical question of what the right decision theory is, the goal of AI poses the question of how to implement any given decision theory in an AI. For example, how would one go about building an AI whose behavior matches evidential decision theory’s recommendations? Conversely, we can ask which decision theories (if any) describe the behavior of any existing AI design. In this paper, we study what decision theory an approval-directed agent, i.e., an agent whose goal it is to maximize the score it receives from an overseer, implements. If we assume that the overseer rewards the agent based on the expected value of some von Neumann–Morgenstern utility function, then such an approval-directed agent is guided by two decision theories: the one used by the agent to decide which action to choose in order to maximize the reward and the one used by the overseer to compute the expected utility of a chosen action. We show which of these two decision theories describes the agent’s behavior in which situations.

     

The perceptual origins of the abstract same/different concept in human infants

Very few experiments have studied the two item same/different relation in young human infants. This contrasts with an extensive animal literature. We tested young infants with two novel tasks designed specifically to provide convergent comparative measures. Each infant completed both tasks allowing an assessment of their understanding of the abstract concept rather than task-specific abilities. In a looking time task with photographic stimuli, we found that 8-month-olds are sensitive to the relation but 4-month-olds are not. The second task used an anticipatory eye movement paradigm with simple geometric stimuli. On each trial, two colored shapes appear and moved upwards behind an occluder. They reappeared on either the upper left or right depending on the relation between them. Infants at both ages learned and generalized the dependency but only for the different relation. These results show that human infants can learn the same/different concept but that, in strong continuity with animal results, their abilities are firmly grounded in perception.     

Internet-based treatment of depression: a randomized controlled trial comparing guided with unguided self-help

Internet-delivered self-help for depression with therapist guidance has shown efficacy in several trials. Results from meta-analyses suggest that guidance is important and that self-help programs without support are less effective. However, there are no direct experimental comparisons between guided and unguided internet-based treatments for depression. The present study compared the benefits of a 10-week web-based unguided self-help treatment with the same intervention complemented with weekly therapist support via e-mail. A waiting-list control group was also included. Seventy-six individuals meeting the diagnostic criteria of major depression or dysthymia were randomly assigned to one of the three conditions. The Beck Depression Inventory (BDI-II) was used as the primary outcome measure. Secondary outcomes included general psychopathology, interpersonal problems, and quality of life. Sixty-nine participants (91%) completed the assessment at posttreatment and 59 (78%) at 6-month follow-up. Results showed significant symptom reductions in both treatment groups compared to the waiting-list control group. At posttreatment, between-group effect sizes on the BDI-II were d = .66 for unguided self-help versus waiting-list and d = 1.14 for guided self-help versus waiting-list controls. In the comparison of the two active treatments, small-to-moderate, but not statistically significant effects in favor of the guided condition were found on all measured dimensions. In both groups, treatment gains were maintained at 6-month follow-up. The findings provide evidence that internet-delivered treatments for depression can be effective whether support is added or not. However, all participants were interviewed in a structured diagnostic telephone interview before inclusion, which prohibits conclusions regarding unguided treatments that are without any human contact.     

The ethics of morphing

Here’s one piece of practical reasoning: “If I do this then a person will reap some benefits and suffer some costs. On balance, the benefits outweigh the costs. So I ought to do it.” Here’s another: “If I do this then one person will reap some benefits and another will suffer some costs. On balance, the benefits to the one person outweigh the costs to the other. So I ought to do it.” Many influential philosophers say that there is something dubious about the second piece of reasoning. They say that it makes sense to trade-off costs and benefits within lives, but not across lives. In this paper I make a case for the second piece of reasoning. My case turns on the existence of morphing sequences—sequences of possible states of affairs across which people transform smoothly into other people.

Tracking the implicit self using event-related potentials

Negative biases in implicit self-evaluation are thought to be detrimental to subjective well-being and have been linked to various psychological disorders, including depression. An understanding of the neural processes underlying implicit self-evaluation in healthy subjects could provide a basis for the investigation of negative biases in depressed patients, the development of differential psychotherapeutic interventions, and the estimation of relapse risk in remitted patients. We thus studied the brain processes linked to implicit self-evaluation in 25 healthy subjects using event-related potential (ERP) recording during a self-relevant Implicit Association Test (sIAT). Consistent with a positive implicit self-evaluation in healthy subjects, they responded significantly faster to the congruent (self–positive mapping) than to the incongruent sIAT condition (self–negative mapping). Our main finding was a topographical ERP difference in a time window between 600 and 700 ms, whereas no significant differences between congruent and incongruent conditions were observed in earlier time windows. This suggests that biases in implicit self-evaluation are reflected only indirectly, in the additional recruitment of control processes needed to override the positive implicit self-evaluation of healthy subjects in the incongruent sIAT condition. Brain activations linked to these control processes can thus serve as an indirect measure for estimating biases in implicit self-evaluation. The sIAT paradigm, combined with ERP, could therefore permit the tracking of the neural processes underlying implicit self-evaluation in depressed patients during psychotherapy.     

Internet-based treatment of social phobia: a randomized controlled trial comparing unguided with two types of guided self-help

Internet-based self-help for social phobia with minimal therapist support via email have shown efficacy in several controlled trials by independent research teams. The role and necessity of therapist guidance is, however, still largely unclear. The present study compared the benefits of a 10-week web-based unguided self-help treatment for social phobia with the same intervention complemented with minimal, although weekly, therapist support via email. Further, a third treatment arm was included, in which the level of support was flexibly stepped up, from no support to email or telephone contact, on demand of the participants. Eighty-one individuals meeting diagnostic criteria for social phobia were randomly assigned to one of the three conditions. Primary outcome measures were self-report measures of symptoms of social phobia. Secondary outcome measures included symptoms of depression, interpersonal problems, and general symptomatology. Measures were taken at baseline, post-treatment, and at 6-month follow-up. Data from a telephone-administered diagnostic interview conducted at post-treatment were also included. Results showed significant symptom reductions in all three treatment groups with large effect sizes for primary social phobia measures (Cohen’s d = 1.47) and for secondary outcome measures (d = 1.16). No substantial and significant between-groups effects were found on any of the measures (Cohen’s d = 00–.36). Moreover, no difference between the three conditions was found regarding diagnosis-free status, clinically significant change, dropout rates, or adherence measures such as lessons or exercises completed. These findings indicate that Internet-delivered treatment for social phobia is a promising treatment option, whether no support is provided or with two different types of therapist guidance.     

Prospective effects of emotion-regulation skills on emotional adjustment

Deficits in emotion-regulation skills have widely been shown to be associated with poor emotional adjustment. However, it is still unclear whether these deficits are a cause or a consequence of poor adjustment. The purpose of the present research was to clarify the reciprocal effects between these 2 concepts. In 2 studies (Ns = 446 and 635), self-reports of emotion regulation and emotional adjustment were assessed twice with a 2-week interval. Cross-lagged regression analyses demonstrated that self-reports of emotion regulation predicted subsequent adjustment, over and above the effects of previous adjustment, whereas emotional adjustment did not predict subsequent emotion regulation. Thus, a focus on emotion-regulation skills may be important in the prevention and treatment of affect-related mental health problems. (PsycINFO Database Record (c) 2010 APA, all rights reserved).