Computational Inference in Cognitive Science Operational Societal and Ethical Considerations

2025-04-29 0 0 351.02KB 9 页 10玖币

侵权投诉

COMPUTATIONAL INFERENCE IN COGNITIVE SCIENCE:

OPERATIONAL, SOCIETAL AND ETHICAL CONSIDERATIONS

TECHNICAL REPORT

Baihan Lin

Department of Neuroscience

Columbia University

New York, NY 10027

baihan.lin@columbia.edu

October 26, 2022

ABSTRACT

Emerging research frontiers and computational advances have gradually transformed cognitive science

into a multidisciplinary and data-driven ﬁeld. As a result, there is a proliferation of cognitive theories

investigated and interpreted from different academic lens and in different levels of abstraction. We

formulate this applied aspect of this challenge as the computational cognitive inference, and describe

the major routes of computational approaches. To balance the potential optimism alongside the

speed and scale of the data-driven era of cognitive science, we propose to inspect this trend in more

empirical terms by identifying the operational challenges, societal impacts and ethical guidelines in

conducting research and interpreting results from the computational inference in cognitive science.

Keywords Cognitive science ·Computational inference ·Ethics ·Society ·Digital health ·Machine learning

1 Introduction

There is a research trend in cognitive science that shifts from a top-down direction (guided by hypothesis-driven testing

of cognitive theories) towards a bottom-up approach (enabled by data-drivcen pattern discovery of cognition-related

properties). The emergence of high-throughput data collection techniques provides cognitive scientists rich research

substances of labelled behavioral data, from one’s digital traces on a social media, to large-scale crowdsourcing of

experimental responses to well-deﬁned cognitive tasks [

]. Riding along the big data era of cognitive science is the

advanced developments of artiﬁcial intelligence (AI) methods that is capable of performing components of cognitive

functions at human-level or superhuman-level performances. With the new directions, comes new challenges. As

the study of the essence, tasks and functions of cognition, how can we as cognitive scientists reshape the ﬁeld using

these new sources of data and new tools of analytical methods, such that it maintains a coherent core as the classical

theory-driven studies of cognitive science?

To better formulate this challenge, we categorizes the interactions between the concepts of AI and those of the human

cognition into three main types (Figure 1). First, we have the computational inference, the process of utilizing machine

learning models as a prediction or inference engines to map from measurable signals to the cognitive properties. The

second direction is to use the cognitive theory as a prior to build AI. This approach can be dated as early as the symbolic

cognitive architectures in 1970s [

], where major cognitive processes such as knowledge representation, memory,

learning and control are explicitly mapped into computational components. A recent perspective piece by [

] further

points out the missing pieces in this direction: causality, intuition, compositionality and generalizability. The third

direction, computational modeling, is to construct brain-computational models that mimic the certain properties of

cognition or neurobiology and then compare them against experimental data. This approach is also termed as the

cognitive computational neuroscience by [

] which proposes to use task-performing computational models to test the

cognitive processes against implementation-level hypotheses of neurobiologically plausible dynamic components, as

well the main considerations in this interdisciplinary approach. Since the empirical considerations of the last two

arXiv:2210.13526v1 [q-bio.NC] 24 Oct 2022

Computational Inference in Cognitive Science: Operational, Societal and Ethical Considerations

Figure 1: The three interactions between the concepts of AI and those of the human cognition

approaches has been discussed in the above work, we will focus our discussion to the ﬁrst direction, the computational

inference problem in cognitive science.

A simple distinction of the three approaches is that: the computational modeling of the mind refers to the use of

computers to simulate the workings of the human mind; the symbolic reasoning of the mind is the ability of the mind to

reason using symbols and abstractions; and the computational inference of the mind is the ability of the mind to make

inferences based on computation. One might relate this categorization to the Marr’s distinction of the three levels of

analysis: the computational theory, the algorithm, and the neurobiological or physical implementation [

]. The algorithm

level maps to building symbolic AI systems. The neurobiological implementation level maps to modeling computational

mechanisms using biologically plausible model components. The objective of the computational inference, on the other

hand, is to ﬁnd the best surrogate models to predict the components and sub-processes of the computational theory.

The top-down perspective of cognitive science decomposes complex cognitive processes into simplier subprocesses

which has their own computational components. However, understanding these subprocesses (e.g. control, perception,

learning and decision making) doesn’t warrant a uniﬁed and coherent theory of cognition until very recently [

]. Unlike

the nascent development of ﬁnding a uniﬁed thoery in the basic research, the machine learning models which are

independently tailored for speciﬁc inference tasks can already be readily used for applied research. As George Box the

statistician wisely points out, “all models are wrong, but some are useful” [

]. Here we argue that, the computational

cognitive inference is the cornerstone of applied cognitive science because they provide actionable inference anchors to

cognitive concepts which can be efﬁciently binded with important downstream real-world applications and if taken

responsible discretion, provide useful interpretable insights in clinical setting.

In this work, we aim to address the conceptual challenges of inferring properties of cognition with computational models,

in particular, from an applied point of view. First, we will formulate the research problem, the measurement approaches,

and major routes of computational methods. What are the operational challenges of developing a data-driven cognitive

science study? What are the societal impacts along with the analyses and interpretations of the experimental results?

And what are the ethical guidelines to design, perform and analyze study of computational cognitive inference?

2 Problem Setting

Here we ﬁrst formulate the problem setting. From a statistical and signal processing point of view, the cognition is an

unknown data generating process in the brain that “generates” the data, which we record externally with behavioral,

imaging and physiological measurements. These measurements serve as an approximation to the underlying cognitive

mechanism. The goal of computational inference is to ﬁnd the most accurate computational account of the measured

data to reﬂect the underlying cognitive properties.

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

10 玖币 0人已下载

立即下载

摘要：

COMPUTATIONALINFERENCEINCOGNITIVESCIENCE:OPERATIONAL,SOCIETALANDETHICALCONSIDERATIONSTECHNICALREPORTBaihanLinDepartmentofNeuroscienceColumbiaUniversityNewYork,NY10027baihan.lin@columbia.eduOctober26,2022ABSTRACTEmergingresearchfrontiersandcomputationaladvanceshavegraduallytransformedcognitivescience...

展开>> 收起<<

Computational Inference in Cognitive Science Operational Societal and Ethical Considerations.pdf

共9页,预览2页

还剩页未读，继续阅读

声明：本站为文档C2C交易模式，即用户上传的文档直接被用户下载，本站只是中间服务平台，本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私，请立即通知玖贝云文库，我们立即给予删除！

Computational Inference in Cognitive Science Operational Societal and Ethical Considerations

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

热门标签

举报选择: