The self as a generative, teleological, and subjective prior: Mutually-modulated temporal agency
The self as a generative, teleological, and subjective prior: Mutually-modulated temporal agency
复制标题
DOI：
10.1101/519934
复制
发表时间：
2019-01
期刊：
影响因子：
--
通讯作者：
Tomohisa Asai;Shu Imaizumi;H. Imamizu
中科院分区：
文献类型：
--
作者：
Tomohisa Asai;Shu Imaizumi;H. Imamizu
我们如何证明自己的存在？最近的理论研究表明，在大脑中，结果反过来推断出自我作为原因存在的可能性。虽然贝叶斯关于代理意识的观点被广泛接受，但支持这一理论的经验证据仍然缺乏。目前的研究在7个实验中考察了时间感知方面的结果调节代理，共有90名参与者。假设知觉的产生，而不是终止，应该主观地推断自我的存在，即使两者包括相同的刺激，并由相同的目的论行为驱动。结果支持假设（实验1和2）。与由相同的意志按键驱动的终止相比，参与者判断出更强的自我能动性，检测到更少的延迟，或者感觉听觉产生的持续时间更短。此外，主要实验（实验3）关注的是行为和结果的时间概率分布（如标准差或相对熵），并得出结论，在开始/抵消代理中观察到的对比表明，只有在代理行为中，因果之间才存在相互或双向关系，其特征是主动（即目的性）产生。最后，基于能动性的三个维度（生成性、目的性和主观性），讨论了意志动作调节时间感知的并行理论模型。代理感是指产生自己行为的主观体验（Gallagher, 2000）。在过去的十年里，对这一主题的研究有所增加，因为对人类代理的理解的增加可能是了解我们如何意识到自己的一个窗口，并可能使我们更好地了解精神分裂症等精神疾病。然而，“自由意志”、“自我意识”和/或“意志”是很难进行实证调查的概念（Frith & Haggard, 2018）。研究人员并不确切地知道在“代理行为”期间我们的感官发生了什么。然而，三个概念可以代表代理的真正维度：生成性、目的性和主观性（Haggard, 2019）。最近的理论研究表明，在大脑中，结果反过来推断出自我作为原因存在的可能性。贝叶斯关于能动性的观点可以为一些经典命题提供解释，比如“ ego cogito, ergo sum（我思故我在）”，“ cogito cogito, ergo cogito sum（我思故我在）”，“我动故我在”或“我预测故我在”。（例如，Corlett, 2017）。所有这些理论主张都隐含地涉及贝叶斯“逆”概率或推理（例如，“思想即行动”），其中效果（感知）追溯地追溯了原因（行动），以及预测后续感知的行动（图1，包括生成、目的论和主观性作为代理的三重维度）。行动和感知之间的这种双向时间关系可以根据“自我证明”或“先验自我”进行经验检验。图1所示。动作和感觉在时间上相互依赖的理论假设。注意：贝叶斯主观概率忽略了因果之间的顺序（即物理时间箭头），即使我们区分了两者（预测vs后置）。此外，动作计时和感觉计时的后验分布（由于消除了前/后置误差）是相互牵拉的，因为两者具有相同的先验（即，正如当前研究表明的那样，自我必须是直接的主动发生器）。图标由Freepik从www.flaticon.com制作。目前的研究在时间感知方面对效应调节代理进行了七个实验，共有90名参与者。我们假设知觉的产生，而不是终止，应该反向推断行动中的自我，即使两者包括相同的刺激，并由相同的行动驱动。结果支持这一假设（实验1和2）。参与者判断更强的自我代理，检测到更少的延迟，或者感觉听觉产生的持续时间更短，与终止相比，在由他/她自己的意志驱动的情况下，但也可能有不同的“先验”（即，自我或他人，作为绝对因果结构选择或模型加权）。此外，主要实验实验3侧重于同一时间轴上行为和结果的概率分布（例如标准差或相对熵），并得出结论，这种开始/抵消代理的对比表明，仅在代理行为中，因果之间存在相互或双向关系，其特征是主动生成（图2）。总之，结果表明，自我是作为一个独立的主体学习的，它可以立即通过自己的意志在环境中引起行动。这种解释与代理感的字面含义以及代理的三重维度理论高度一致。图2。总结了实验1（A，B，C）和2（A，B，C）的结果。注：星号仅表示起始条件和偏移条件之间的显著差异。虚线拱表示on/offset条件的主要影响。误差条代表±1se。** p < .01, * p < .05，†p < .10，完整方差分析结果见表S2。我们认为，目前的研究成功地观察到以前不可见的行为和感知之间的关系，有利于基于信息理论、计算神经科学（图3）和理论讨论的高级分析。图3。实验3的动作、音调和意向结合效应。注：虚线拱为主要影响因素。误差条代表±1se。** p < .01, * p < .05，†p < .10，完整方差分析结果见表S2。本手稿未在其他地方部分或全部发表或呈现。没有需要申报的利益冲突。我预测，因此我在：氯胺酮和精神分裂症下的紊乱预测编码。生物精神病学杂志，81,465-466（2017）。弗里思，c.d.，哈格德，P.《意志与大脑——重新审视一个经典的实验研究》。神经科学进展，41,405-407（2018）。自我的哲学概念：对认知科学的启示。Cogn趋势。科学通报，2004,14 - 21（2000）。《人类意志的神经认知基础》。精神病学年鉴。(新闻)
How can we certify our existence? Recent theoretical studies have suggested that, in the brain, the effect inversely infers the possibility of the existence of the self as the cause. While this Bayesian view of the sense of agency is widely accepted, empirical evidence in support of this theory is still lacking. The current study examined outcome-modulated agency in terms of time perception in seven experiments, with a total of 90 participants. It was hypothesized that perceptual generation, not termination, should subjectively infer the existence of the self, even though both include the same stimuli and are driven by the same teleological action. Results suggest support for the hypothesis (Experiments 1 and 2). Participants judged stronger self-agency, detected less delay, or felt shorter duration for auditory generation, compared with termination, which was driven by the same volitional key-press. Furthermore, the main experiment, Experiment 3, focused on temporal probability distributions both for action and outcome (e.g., standard deviations or relative entropy), and concluded that the observed contrast in onset/offset agency indicates a mutual or bidirectional relationship between cause and effect only during agentive action, characterized by active (i.e., teleological) generation. Finally, the concurrent theoretical models for volitional action-modulated time perception are discussed on the basis of the suggested triplet (generativity, teleology, and subjectivity) dimensions of agency. Extended abstract The sense of agency refers to the subjective experience of generating one’s own actions (Gallagher, 2000). Within the past decade, research on this subject has increased, as increased understanding of human agency could be a window into knowing how we are aware of ourselves, and possibly allowing for a better understanding of mental illnesses, like schizophrenia. However, “free will,” “self-consciousness,” and/or “volition” are concepts that are always difficult to empirically investigate (Frith & Haggard, 2018). Researchers do not know exactly what occurs in our sensorium during “agentive action.” However, three conceptions could represent genuine dimensions of agency: generativity, teleology, and subjectivity (Haggard, 2019). Recent theoretical studies have suggested that, in the brain, the effect inversely infers the possibility of the existence of the self as the cause. This Bayesian view of the sense of agency may provide an explanation for classical statements, such as “ego cogito, ergo sum (I think, therefore I am),” “cogito cogito, ergo cogito sum (I think that I think, therefore I think that I am),” “I move, therefore I am,” or “I predict, therefore I am.” (e.g., Corlett, 2017). All of these theoretical claims implicitly refer to the Bayesian “inverse” probability or inference (e.g., “thought as action”), where the effect (perception) retrospectively postdicts the cause (action), as well as the action predicting the following perception (Figure 1, including generativity, teleology, and subjectivity as triplet dimensions of agency). This bidirectional temporal relationship between action and perception can be empirically examined in terms of the “certification of the self” or self as prior. Figure 1. The theoretical hypothesis of cross-dependency between action and sensation in time. Note: Bayesian subjective probability ignores the order between cause and effect (i.e., the physical temporal arrow), even if we make a distinction between the two (prediction vs. postdiction). Furthermore, the posterior distributions (as a result of dissolving the pre/postdiction error) for action-timing and sensation-timing are mutually-tugged since both share the same prior (i.e., the self must be the immediate active generator, as the current study suggests). Icon made by Freepik from www.flaticon.com. The current study examined effect-modulated agency in terms of time perception in seven experiments, conducted with a total of 90 participants. We hypothesized that perceptual generation, not termination, should inversely infer the self in action, even though both include the same stimuli, and are driven by the same action. Results support this hypothesis (Experiments 1 and 2). Participants judged stronger self-agency, detected less delay, or felt shorter duration for auditory generation, compared to termination, in instances driven by a key-press of his/her own volition, but may have also had different “priors” (i.e., self or other, as a categorical causal structure selection or model weighting). Furthermore, the main experiment, Experiment 3, focused on probability distributions both for action and outcome on the same temporal axis (e.g., standard deviations or relative entropy), and concluded that this contrast in onset/offset agency indicates the mutual or bidirectional relationship between cause and effect only during agentive action, characterized by the active generation (Figure 2). In summary, results indicate that the self is learned as a prior to being an independent agent that can immediately cause actions to occur within the environment merely through one’s own volition. This interpretation is highly congruent with the literal meaning of the sense of agency and also with the theory of the triplet dimensions of agency. Figure 2. Summarized results across Experiments 1(A,B,C) and 2(A,B,C). Note: Asterisks represent the significant differences only between onset and offset conditions. Dashed arches represent the main effect of on/offset condition. Error bars represents ± 1 SE. ** p < .01, * p < .05, † p < .10, see Table S2 for full ANOVA results. We believe that the current study successfully observed the previously invisible relationship between action and perception in favor of advanced analysis, based on information theory, computational neuroscience (Figure 3), and theoretical discussions. Figure 3. The motor, tone, and intentional binding effect in Experiment 3. Note: Dashed arches represent the main effects of factors. Error bars represents ± 1 SE. ** p < .01, * p < .05, † p < .10, see Table S2 for full ANOVA results. This manuscript has not been published or presented elsewhere in part or in entirety. There are no conflicts of interest to declare. References Corlett, P. I Predict, Therefore I Am: Perturbed Predictive Coding Under Ketamine and in Schizophrenia. Biological Psychiatry, 81, 465–466 (2017). Frith, C. D., Haggard, P. Volition and the Brain - Revisiting a Classic Experimental Study. Trends Neurosci. 41, 405–407 (2018). Gallagher, S. Philosophical conceptions of the self: implications for cognitive science. Trends Cogn. Sci. 4, 14–21 (2000). Haggard, P. The Neurocognitive Bases of Human Volition. Annu Rev Psychol. (in press)