Broadening AI Ethics Narratives: An Indic Art View

by   Ajay Divakaran, et al.

Incorporating interdisciplinary perspectives is seen as an essential step towards enhancing artificial intelligence (AI) ethics. In this regard, the field of arts is perceived to play a key role in elucidating diverse historical and cultural narratives, serving as a bridge across research communities. Most of the works that examine the interplay between the field of arts and AI ethics concern digital artworks, largely exploring the potential of computational tools in being able to surface biases in AI systems. In this paper, we investigate a complementary direction–that of uncovering the unique socio-cultural perspectives embedded in human-made art, which in turn, can be valuable in expanding the horizon of AI ethics. Through qualitative interviews of sixteen artists, art scholars, and researchers of diverse Indian art forms like music, sculpture, painting, floor drawings, dance, etc., we explore how non-Western ethical abstractions, methods of learning, and participatory practices observed in Indian arts, one of the most ancient yet perpetual and influential art traditions, can inform the FAccT community. Insights from our study suggest (1) the need for incorporating holistic perspectives (that are informed both by data-driven observations and prior beliefs encapsulating the structural models of the world) in designing ethical AI algorithms, (2) the need for integrating multimodal data formats for design, development, and evaluation of ethical AI systems, (3) the need for viewing AI ethics as a dynamic, cumulative, shared process rather than as a self contained framework to facilitate adaptability without annihilation of values, (4) the need for consistent life-long learning to enhance AI accountability, and (5) the need for identifying ethical commonalities across cultures and infusing the same into AI system design, so as to enhance applicability across geographies.


page 4

page 10


Building Bridges: Generative Artworks to Explore AI Ethics

In recent years, there has been an increased emphasis on understanding a...

Some Critical and Ethical Perspectives on the Empirical Turn of AI Interpretability

We consider two fundamental and related issues currently faced by Artifi...

Why we need biased AI – How including cognitive and ethical machine biases can enhance AI systems

This paper stresses the importance of biases in the field of artificial ...

Identifying Ethical Issues in AI Partners in Human-AI Co-Creation

Human-AI co-creativity involves humans and AI collaborating on a shared ...

Toward Ethical AIED

This paper presents the key conclusions to the forthcoming edited book o...

Social Inclusion in Curated Contexts: Insights from Museum Practices

Artificial intelligence literature suggests that minority and fragile co...

Comprehensiveness of Archives: A Modern AI-enabled Approach to Build Comprehensive Shared Cultural Heritage

Archives play a crucial role in the construction and advancement of soci...

1. Introduction

Incorporating interdisciplinary perspectives is an essential requisite for the design of ethical artificial intelligence (AI) systems (Raji et al., 2021; Romm, 1998; Krafft et al., 2020; Patel et al., 2019)

. From humanities and social sciences to law and policymaking, a growing body of excellent works have elucidated important lessons from diverse fields to the FAccT community. For example, works from moral and political philosophy have been leveraged to highlight emerging debates concerning fair machine learning

(Binns, 2018), to shed new light on human practices by situating technology within the human, social, and the political goals (Bietti, 2020), and to suggest revised approaches in conceptualizing and operationalizing algorithmic recourse (Venkatasubramanian and Alfano, 2020). Studies have examined how document collection practices in archives can inform data collection in machine learning (Jo and Gebru, 2020). Principles from quantitative social sciences have been proposed as frameworks in understanding fairness in computational systems (Jacobs and Wallach, 2021). Lessons from feminist studies have been used to offer guidelines on how to conduct research on explanations (Hancox-Li and Kumar, 2021), to inform data collection process in AI systems (Leavy et al., 2021), and to study issues related to algorithmic oppression (Hampton, 2021). Findings from theories of parenting have been used to inform new ways of designing and developing autonomous agents (croeser and Eckersley, 2019). Issues concerning governance of AI technologies have been discussed leveraging design lessons from history (Cihon et al., 2020).

In addition to the aforementioned diverse disciplines, the field of arts is regarded to play an important role in shaping fairness and accountability of AI systems (Jensen, 2020; Srinivasan and Uchino, 2021). As stated in (Gorichanaz, 2020), art is ‘a form of technology that contributes to knowledge production by exemplifying aspects of the world that would otherwise go overlooked’. It is also believed that art facilitates cultivation of ‘moral knowledge’, i.e. knowledge about what is right and what is not right (Young, 2001). The authors in (Novitz, 1987) state that art also engenders empathy by creating a means of direct experience to compare different viewpoints about the world. Art is regarded as a powerful language that can bring about symbolic transposition (Leroi-Gourhan, 1993), and is perceived to be a form of communication with the public (Daniele and Song, 2019).

Scholars posit that the field of arts has the potential to transcend the limitations and distortions of dominant ethical paradigms derived from the sciences and economics by being able to showcase plurality of values, by elucidating procedures as opposed to just outcomes, and by emphasizing the role of participatory design (Tasioulas, 2021). Researchers argue that artists have historically deployed new technologies in unexpected and often prescient ways (Shaken, 2002) and have been interpreted as vanguards: of new ideas, techniques, and cultural practices (Francastel, 2000; Stark and Crawford, 2019). Furthermore, the authors in (Stark and Crawford, 2019) argue that artists engaging with data analytic tools and computational techniques can share their opinions on the changing and nuanced ethical questions faced by those who use data and AI algorithms in their work. For instance, American artist Rashaad Newsome tells stories of racial injustice through the lens of colonialism to highlight existing shortcomings of AI models (Jensen, Retrieved 2021). Inspired by the ‘theory of new aesthetics’ (Leech, 2011)

, artist Avital Meshi deconstructs the notion of ‘whiteness’ in face recognition algorithms by using performance as a tool

(Meshi, 2021). In ImageNet Roulette, an art project designed by researcher Kate Crawford and artist Trevor Paglen, biases associated with large scale image datasets and their adverse consequences are exposed (Crawford and Paglen, 2019).

A majority of works that examine the interplay between arts and AI ethics concern digital artworks, largely exploring the potential of computational tools in supporting artists to surface issues related to bias and discrimination in AI systems. However, there is little to no work that examines the potential of human-made artworks from non-Western contexts in informing the FAccT community—Non-Western art forms can potentially bring to light several novel and unique ethical abstractions that are relevant in local geographies. Given that most studies concerning AI ethics are influenced heavily by Western cultures in their definitions, axes of discrimination, and philosophical roots (Sambasivan et al., 2021), understanding non-Western socio-cultural contexts becomes even more pertinent (Hanna et al., 2020; Abebe et al., 2021). Through this work, we hope to address this gap by studying how art forms from India (here we consider, in a breadth first fashion, a large subset of Indian arts that is inspired by traditional works),—a non-Western country with diverse and rich cultural history, constituting nearly 17 of the world’s population, and which has the largest diaspora in the world— can inform issues related to AI ethics. In particular, we explore the following research questions (RQs):

  • RQ1: What kind of ethical abstractions can be learned from Indian arts?

  • RQ2: Can Indian arts inform the FAccT community of new forms of learning that could potentially help in enhancing fairness and accountability of AI systems?

  • RQ3: Can Indian arts shed light on new forms of participatory approaches that can be useful in the AI pipeline?

  • RQ4: Can we learn new ways of designing adaptable and robust AI systems by looking at Indian artists and arts?

In order to study the aforementioned research questions, we conduct interviews of artists/practitioners, art scholars, researchers, and art aficionados. Given that information about certain forms of arts, especially performing arts like dance and music, cannot be analyzed purely based on literature reviews (Venugopal, 2018), and more broadly, given that interfacing with practitioners and other domain experts can better illumine aspects related to the process of art information creation, sharing, and distribution (Srinivasan et al., 2021), we adopt an interview based approach. Responses from sixteen interview participants whose background span diverse Indian art forms such as Indian classical music (Carnatic and Hindustani forms), Indian style painting (silk cloth painting, Thanjavur paintings), Indian dance (Kuchipudi and Vilasini dance forms), Kolam (traditional geometric floor drawings) among others, are summarized in the context of the aforementioned RQs.

Insights from our study suggest the need for (1) incorporating holistic perspectives (a combination of bottom-up and top down mechanisms that are informed by both data based observations and prior beliefs that encapsulate structural models of the world) in designing ethical AI algorithms, (2) integrating multimodal data formats (sound signals, ancient texts, images, etc.) in the design, development, and evaluation of ethical AI systems, (3) viewing AI ethics as a dynamic, diverse, cumulative, and shared process rather than as a static, self-contained framework to facilitate adaptability without annihilation of values, (4) consistent life-long learning to enhance AI accountability, and (5) identifying ethical commonalities and infusing the same in AI system design to enhance applicability across geographies.

The rest of the paper is organized as follows. We first provide a brief review of Indian art forms outlining their history, influence, and relevance in the context of this study. We then provide details of the interview process starting from questionnaire structure and design to details pertaining to respondent selection and data consent. We then summarize the interviews responses in the light of the research questions. Next, we outline insights from our study by recommending potential pathways towards enhancing AI ethics. Finally we close by discussing limitations and future work.

2. Understanding Indian arts

Scholars and historians of Indian art have long grappled with the problem of describing and understanding a vast diversity of forms of expression, over vast periods of time and space, motivated by a wide variety of psychological and philosophical traditions. Some have got around this by providing a chronology over time of the evolution and development, while others have tried to identify the chief features of particular schools or styles within each genre (Vatsyayan, 1997).

India has a long history of art and craft (Dhar, 2011). The art of the Indus Valley Civilisation (Bronze Age), which included parts of Northern India, was depicted through pottery, seals, jewellery, sculpture, etc. Most regions of India have preserved their art forms for over 3000 years through informal local communities and guilds. Commonly known as folk art, the art forms of these informal local communities are less structured and more organically imbibed than their classical traditions. Stories based on local idioms are woven into songs, paintings, puppetry, weaving, sculptures, and more. Figure 1 provides an illustration of various forms of folk music and folk dance prevalent across different states of India.

Figure 1. Left: An illustration of the diverse folk dance forms overlaid on the map of India corresponding to their respective states. Right: An illustration of various forms of folk music practiced across different Indian states. Maps included after obtaining permission from source—

These traditions have sustained over time in physical form across myriad architectural, sculptural, and other art forms through canonical texts as well as through oral transmission of specific skills and knowledge over generations. India’s oldest texts - the Vedas - speak of two kinds of knowledge. The Mundaka Upanishad (an ancient Sanskrit Vedic text in the Atharva Veda) talks about ‘apara’ knowledge (superficial knowledge, amenable to linguistic analysis) and ‘para’ (higher knowledge, which is related to consciousness). Traditional sciences that one learns through formal training (known as ‘siksa’) is apara, whereas deeper knowledge that arises out of inner inquiry is para or also referred as ‘vidya’ (Rajgopalan, 2021). The Indian arts, like much of all Indian knowledge systems, have to be approached from this perspective.

Elaborating on the continuum evident in Indian arts, in her essay ‘Inter-relationship of the Arts’ (Vatsyayan, Retrieved 2021), Indian art historian Kapila Vatsyayan writes, “The continuity could be maintained only because the tradition itself had an in-built paradigm of facilitating change, constantly adjusting itself to a contemporaneity of time and place while adhering to certain underlying principles which were perennial and immutable. The dynamics of the still centre and the ever kaleidoscopically changing movements of styles, schools and artistic expressions were sustained in different milieus and at different levels of society through different diverse media. The aesthetics emerged from a shared world-view with the acceptance of diversity in form.” The existence of a common ‘substratum or thread’ across diverse forms is not only recognized within the field of Indian arts, but also acknowledged across Indian Philosophical texts. For example, the Rig Veda (6.47.18) describes this ‘unifying thread’ across diverse forms in the following verse (Kak, 2021):

rupam rupam pratirupo babhuva

tadasya rupam praticaksanaya
The translation of the above Sanskrit verse reads thus—

That became the original form of every form

It is That form which is everywhere to be seen

Reflecting on the spiritual power associated with Indian art forms, in their book ‘The Art of Ancient India - Buddhist, Hindu, Jain’, authors Susan L Huntington and John C Huntington write— “Through the process of viewing an image, the devotee gains spiritual power. Indeed, the emphasis on darsana (i.e. viewing) in the Indic culture, which requires only seeing and being in the presence of something or someone great to gain merit, has lent the visual arts an importance, never exceeded elsewhere in the world” (Huntington and Huntington, 2016).

Moving away from a chronological or even a style-based study of art, there has been an attempt by modern artists and scholars to look at Indian art as a manifestation of the ‘inner self’ or the common substratum referred earlier, through external symbols, motifs, and sounds. In his book ‘Towards Ananda - Rethinking Indian Art and Aesthetics’, author Shakti Maira points out two reasons as to why it is important to integrate Indian arts into the global landscape, which would also include fields like AI. The first is that Indic culture has preserved art in the lives of the common man. The second point he makes is that Indian aesthetic traditions, like its Vedantic and Buddhist traditions, have dug deep into the human mind and into the nature of human experience and that “they offer a different and important vision and purpose for art and its practices” (Maira, 2006).

The current dominant philosophy of Postmodernism postulates that all languages and art styles are cultural constructs conditioned by prevailing social circumstances and there are no pure meanings and no one reality (The-Art-Story, Retrieved November 2021). Author Shakti Maira counters Postmodern epistemologists arguing that it is not just the language and art styles that define experiences. He goes so far to say that “While the human experiences that Indian artists refer to are universal, yet there is an ‘Indianness’ in the focus and stress placed on some experiences in preference to others. For example, the emphasis on inner experiences - emotional states and movement of consciousness over sensory experiences and contact with the outer world” (Maira, 2006).

The aforementioned emphasis on inner experience is evident across works concerning Indian aesthetics written in the context of art-making. The concept of ‘Rasa’

or aesthetic flavor was introduced by Sage Bharata (believed to be the father of Indian theatrical forms, who lived roughly between 200 BCE and 200 CE, but estimates vary between 500 BCE and 500 CE) in his seminal work the

‘Natya Shastra’, a Sanskrit treatise on performing arts. According to the Natyashastra, there are nine principal emotional states, human feelings, or moods - shringara (i.e. beautiful), hasya (i.e. humorous or comic), karuna (i.e. compassion, empathetic) raudra (i.e. anger), vira (i.e. brave, heroic), bhayanaka (i.e. fearful), bibhatsa (i.e. odious), adbhuta (i.e. astonishing) and shanta (i.e. peaceful). In the 11 century, Abhinavagupta, a famous theoretician on aesthetics, built a comprehensive theory of feeling in which art and the aesthetic experiences were postulated as an outpouring of energy, creativity, and imagination of the artists and their rasikas (i.e. audience) (Higgins, 2008).

The Indian way of looking inwards is brought out by renowned theatre scholar Richard Schechner in his article on Rasaesthetics (Schechner, 2001), where he mentions the difference between Stanislavsky-based Euro-American acting and the Natya Shastra. Schechner states that in Euro-American acting, “one does not play an emotion ; one plays the ‘given circumstances’, the ‘objectives’, the ‘through-line of action,’ the ‘magic if’. If this is done right, ‘real’ feelings will be experienced and ‘natural’ emotions will be displayed”. Thus, in Western theatre, both American and European, the plot and the text are primary, and it is believed that the story automatically creates the required emotions. On the contrary, across Indian theatre and all forms of performing arts, the emphasis is on emotions with the plot being secondary. According to Schechner’s interpretation of the Natya Shastra’s rasic system, one can work directly on the emotions— “mixing them according to ‘recipes’ known to the great acting gurus (i.e., teachers) —or even by devising new recipes. From a Stanislavskian vantage, such direct work on the emotions will result in false or mechanical acting. But anyone who has seen performers thoroughly trained in the Natya Shastra’s rasic system knows these performers are every bit as effective as performers trained in the Stanislavsky system” (Schechner, 2001). In Schechner’s study of Aristotle’s Poetics and the Natya Shastra, he says that many Indian artists subscribe to the ideal of a theatre that integrates drama, dance, and music. Traditional Indian theatrical genres accomplish this integration in ways that privilege dance, gesture, and music over the plot, thereby emphasizing the associated emotions along with other relevant cognitive and aesthetic elements (Schechner, 2001).

As in theatre, so also in sculpture and other art forms. In her essay (Vatsyayan, Retrieved 2021), Vatsyayan says that Indian sculpture too manifests “this vision of wholeness through a methodology of impersonalisation. Indian figurative art is not portraiture of the specific… Buddha is Buddha, the historical prince Siddhartha, and Sakyamuni, but he is more: he is compassion, pathos and grace in absolute. The spirit and soul of the cosmic infinite is contained in the body of the particular but impersonal form. The image is not the historical figure -it is and it is not the cultural specific in which it is articulated: a Kushan, Gupta, Pala or for that matter an Indonesian, Cambodian, a Khmer, Japanese and Chinese Buddha can be clearly identifiable”

The spiritual and the aesthetic go hand in hand in India. In the book ‘Rasa: Performing the Divine in India’, author Susan Schwartz writes—“It is more fruitful to state that the goal of the aestheticians, from Bharata-Muni onward, has been to facilitate a transformation—of the artist, the audience, and ultimately the world—that may only be understandable from the perspective of religion. So central has the religious context been to understanding and achieving the goals of performance that it is possible to study the religions of India through her performing arts” (Schwartz, 2004).

In a similar vein, author D David Hudson writing on the Vaikunta Perumal Kovil, (ca 770 C.E.) a shrine in the city of Kanchipuram in South India, quotes a poem describing the three postures of the Lord at the three floored sanctum of the temple - the uppermost sanctum is standing, corresponding to the gross material body (referred to as Sthula sharira), the middle sanctum is reclining - the Lord’s subtle material body (referred to as Sukshma sharira) and the third is sitting - the Lord’s pure being (referred to as Suddha Sattva). Apart from its architectural splendor, this shrine signifies a deeper spiritual meaning—that the devotee has to find a means to transition from the state of gross material body (depicted by the standing posture) to the supreme state of bliss without getting trapped in the middle level that corresponds to the endless cycle of life and births (Hudson, 2008). The book also studies the link between this South Indian temple and the Angkor Wat shrine in Cambodia, the largest shrine in the world for an Indian deity.

One of the greatest scholars of Indian art history Ananda Coomaraswamy (1877-1947) in his article ‘The Aims of Indian Art’ (Coomaraswamy, 1975) says that there is a close analogy between the aims of art and theoretical science—“The imagination is required for both; both illustrate that natural tendency to seek the one in the many, to formulate natural laws, which is expressed in the saying that the human mind functions naturally towards unity. The aim of the trained scientific or artistic imagination is to conceive (concipio, lay hold of) invent (invenio, to light upon) or imagine (visualize) some unifying truth previously unsuspected or forgotten. … Ideal art is thus rather a spiritual discovery than a creation. It differs from science in its concern primarily with subjective things, things as they are for us, rather than in themselves. But both art and science have the common aim of unity; of formulating natural laws.” The synthesis of Indian thought is thus understanding the one across many.

Given this broad outline, we have tried to capture the essence of an ancient artistic ethos by interviewing artists and scholars who are engaging with a wide variety of Indian artistic traditions. There is a common function among various art Indian art forms, which is to transform both the artist and the audience using the art form as the vehicle to experience such transcendence. We posit that this functional aspect can be valuable in the context of AI ethics. The section below describes our interview methodology used to study Indian arts, covering both its form and function.

3. Interview Methodology

In order to study the research questions, we adopt interview based analysis. The rationale behind conducting interview based analysis is as follows. First, in the Indian context, the transmission of knowledge in certain forms of arts, especially the performing arts such as dance, music, and theatre, is largely oral. Thus, one can understand the pedagogical foundations by interviewing artists who have worked with generations of masters in the hallowed ‘Guru-Shishya parampara’ (i.e. teacher-student traditions) where information is passed on from the preceptor to the pupil. Second, scholars premise that there are gaps in existing literature on documenting different aspects of development of the Indian arts (Venugopal, 2018). A combination of interviews coupled with literature review therefore helps in gathering a more comprehensive picture of the Indian art scene.

We conducted interviews of a variety of people engaged in different forms of Indian art. These included artists, art history scholars, AI researchers who have trained in an Indian art form, and cross-disciplinary researchers who have studied topics at the intersection of the arts and the sciences. The respondent pool that we reached out to was determined so as to be representative of 1) gender, 2) age, 3) years of practice, 4) the art form, and 5) the region in India they hail from (which in turn has a bearing on training and aesthetics). We also contacted people who did not have Indian origins, but who had extensively studied an Indian art form. All the potential respondents were contacted via email and through their social media profile. In total over thirty people were contacted, out of which we received response from sixteen. The respondents who participated in the interviews include vocalists across different forms of Indian music (Carnatic classical and Hindustani), instrumentalists (Indian instruments Mridangam, Sitar, Sarod, Veena), sculptors, painters (silk cloth painting and Thanjavur painting style), dancers (Kuchipudi and Vilasini Indian dance forms), Kolam artists (a traditional form of geometric floor drawings), scholars of Indic knowledge systems, and computer science/AI researchers. Respondents were distributed across continents–North America (USA), Europe (Sweden and Netherlands), and Asia (various parts of India).

Participation in the interview was voluntary. The respondents were informed of the interview’s purpose; they were provided with some context and illustrations highlighting the need for fairness and accountability in AI systems. The respondents who participated in the study consented to sharing the information provided by them in an academic paper or journal. Respondents were asked some personal information such as their name, location, and email so that we could contact them if we needed any clarification regarding their response. Respondents were also asked to indicate their engagement with arts–as art scholars, art researchers, practitioners/artists, aficionados, etc. We list the names of respondents under acknowledgements, names of respondents who wished to remain anonymous are not included.

The interviews were journalistic in nature - exploratory, topical, cross examining conventional premises (Miller - Carpenter et al., 2018; Svyetlov, 2001). The questions spanned areas that can described as art aesthetics, art education, art history, art psychology, and artificial intelligence. Although the questions were primarily motivated by the research questions, they were designed to be personal, introspective, and open-ended, thereby allowing flexibility in interpretation and adaptation. There is an element of subjectivity to the method as close-ended research cannot capture the highly subjective, creative world of artists. While the pedagogy of the Indic artistic tradition is extremely principled and replicable, the artist adds their own ‘manodharma (improvisation) which makes every artistic work unique.

First, leveraging our backgrounds across fields like Indian art history, music, social science, and artificial intelligence, we designed a set of questions to understand how the learning, training, execution, and reception of an Indian art form, which happens over a long period of time, can be used to address issues related to AI ethics. We further refined the questions and added new questions after consulting scholars of Indian arts and artificial intelligence. We provide a brief description of the question topics below.

In order to understand how people across different disciplines perceive artificial intelligence, we included a question that asked the respondents to explain what they think AI is and list the domains that have influenced their understanding of AI. Some artists and art scholars whom we contacted but who did not participate in the interview cited that they were not too familiar with AI related aspects to participate; thus including this question was mutually beneficial- it helped the respondents to determine their participation, and helped us in obtaining informed responses.

We included specific questions pertaining to the art traditions and learning to understand if the process of traditional art information/knowledge sharing can inform the AI community of new forms of learning and representations. We also included questions related to collaborations between artists, how artists connect with the audience, how they are able to achieve harmony in renditions, and how they adapt to changing times to understand if such artistic practices can shed light on new forms of participatory approaches for AI systems. In order to understand if Indian arts can inform the AI community of new ethical concepts, especially those that are relevant in non-Western contexts, we also asked the respondents to describe such Indian ethical concepts which they thought would benefit the AI community. Finally, we also asked if the respondents had any message for AI researchers.

To allow the respondents to respond at their own convenience and time, the interviews were primarily delivered as a questionnaire using Google forms. All participants were asked the same questions in the same manner. The most common response to the questionnaire was that some questions were rather deep, requiring long response times and thought. One respondent instead preferred to discuss over a call. The interview which was conducted via call was manually transcribed to obtain the data for analysis.

4. Summary of Interview Responses

In this section, we provide a summary of the interview responses in the light of the RQs considered for the study, along with our brief extrapolation of the viewpoints suggested in the responses.

4.1. Discussion of RQ1

What kind of ethical abstractions can be learned from the Indian arts?

In order to examine this RQ, in our interview questionnaire, we had included a question asking the respondents to describe, if any, ethical abstractions or concepts from Indian arts that potentially could aid in designing ethical AI systems. We had also included an open ended question that asked the respondents to share their opinions/recommendations with respect to enhancing ethics in AI systems. Below, we summarize these responses.

In response to the question concerning ethical abstractions or concepts from Indian arts that could be useful in design of ethical AI systems, one respondent wrote that the spirit of seeing the universe as one family or the principle of Vishwa Kutumba that is exemplified in Indian philosophy and arts could be valuable. Another respondent mentioned that the principle of ‘Paropakaraya punyaya payapa para peedanam’, which roughly translates to ‘ensuring welfare for all and penalizing immoral acts’ could be valuable for the design of ethical AI systems.

Yet another respondent mentioned that unlike Western philosophies which are largely driven by rule based frameworks and which are based on the premise that individuals are different, Indian philosophy recognizes the commonalities across people, and that this ideology is reflected in the Indian arts as well. According to this respondent, ethical is anything that evokes the inner reality within an individual whereas unethical is that which revolts this process. The respondent further added that the Indian art depicts the common thread across all life forms. This spirit has been extensively reflected across works of many Indian poets and artists who emphasized that the physical basis of all life forms is the same, independent of factors such as gender, religion, caste, and so on (Venkat, 2010).

Echoing the aforementioned thought, another respondent pointed out that Indian art goes beyond binaries and cited the example of ‘Ardhanarishvara’ (a Sanskrit word meaning ’Lord who is half Woman’) which is a composite male-female figure in the Indian sacred pantheon. The respondent added that across many Indian and Southeast Asian sculptures, icons are decorated combining the male and female aspects, which are perceived to be inseparable. In a similar vein, contrary to Western notions of race wherein fair complexion (white) is considered to belong to an advantaged group, poets of ancient India considered dark complexion as something vibrant and throbbing with life and exuberance. For example, almost all of the female characters in Kalidasa’s (considered to be one of India’s greatest playwrights and dramatists) masterpieces were dark complexioned (Paul, 2015).

The aforementioned viewpoints suggest that AI systems should strive for welfare of all stakeholders involved in the AI pipeline, and recognize the commonalities across individuals in decision making, Further, the aforementioned responses emphasize the need look beyond rule based abstractions for characterizing ethics in AI systems. In fact, ancient Indian texts advocate several rules for the observance of justice. These include aspects such as compassion, forgiveness, patience, truthfulness, absence of anger, piety, sanctity, non-violence, and control of senses. These aspects have been illustrated across different art forms. As also pointed out in (Sambasivan et al., 2021), notions beyond fairness such as restorative justice (Boyes-Watson, 2014; Inayatullah, 2006) and those based on moral foundations such as purity and compassion (Graham et al., 2013) can help shed light on philosophical viewpoints that are relevant in local cultural and geographical contexts.

4.2. Discussion of RQ2

Can the Indian arts inform the FAccT community of new forms of learning that could potentially help in enhancing fairness and accountability of AI systems?

In order to analyze this RQ, we asked the respondents to describe the process of information gathering, learning, and sharing in their art forms. The respondents were also asked to explain what was distinctive in their art form and learning process compared to other art forms and learning practices. We also asked them to describe the checks and balances inbuilt in the training and delivery of their art forms, and to describe how they maintain consistency with traditional guidelines and practices. These responses are summarized below.

In response to a question that asked the respondents to describe the process of information gathering and sharing in their art form, one respondent said that there are three sources for artistic endeavor - (1) the words of the guru or the teacher (which emphasises humility, a relationship of shared learning, and gratitude), (2) the textual source (grounding the knowledge of the performing arts in scholarship and tradition, as well as stating the parameters or boundaries of individual versus group ideas and expression), and (3) the artist’s own experience (accentuating vulnerabilities, allowing space for improvisation and articulation of individual inspiration ‘manodharma’, engaging with fellow artists, etc).

Referring to the Indian dance system or Natya shastra, another respondent mentioned that there is an established methodology to understand, codify, and express emotions in Indian dance forms through a five-tier structure of emotions which includes Vibhava or the determinant, Anubhava or the consequent, Sthayi Bhava or the dominant emotion, Vyabhichari Bhava or the transient emotions, and Sattvika Bhava or natural bodily responses to emotion. Apart from engaging the body, Natya is seen as a comprehensive art, fusing several disciplines including music, literature, sculpture, painting, even architecture, and chikitsa shastra (i.e. study of health).

Contrasting the process of human learning with machine learning, another respondent said that the problem about new technologies is that they accord more importance to a certain kind of intelligence over other considerations and added that AI systems try to magnify this in order create a category of ‘superhumans’. The respondent added that AI system emphasize the higher faculties of humans, often ignoring their ‘vulnerability and finitude’ which have been the basis and source of much of human creativity. In response to the question about what is distinctive in their art style and art philosophy, the respondent said that there was a deep connect between the meaning of art and everyday life - which is manifested in performative traditions even today. The respondent added that Indian art has avoided the pitfall of art becoming elitist or merely an intellectual exercise by connecting it to the daily lives of people.

Indian art and philosophy offers different interpretations of intelligence, which have over the last few decades gained acceptance in modern psychology (Singh, 2012). In the Indian view, intelligence refers to ‘waking up, noticing, recognizing, understanding, and comprehending. Buddhi or intellect includes such things as determination, mental effort, feelings, and opinions in addition to intellectual processes such as knowledge, discrimination, and decision making (Singh, 2012). The respondent added that Indian music not only demonstrates the kind of intelligence seen in AI systems— such as higher calculative ability (in Indian rhythm), superlative memory (evidence includes learning of thousands of songs through oral transmission), decision making capacity (while innovating on the spot), high speeds of action (in responding to co-artists), using the prosodic structure well— but it also demonstrates other capabilities often ignored by AI systems. These forms of intelligence have been described as prathibha (keen intellect), prajna (transcendental wisdom), Vak (speech), and Bhavas (emotions).

The Indian arts offer a multi-dimensional image of the human mind. Studying the basic guidelines and methodologies of the making of Indian art, said one respondent, might be valuable in incorporating its core ideas to other disciplines, including AI. Giving the example of Kolams which are sacred geometrical drawings, one respondent outlined how women across India clear the front of their houses and draw Kolams —geometrical pattern of dots and lines— using rice or rock powder to energize the house. It is an ancient artistic ritual which while being temporal (as it can be easily erased) has also remained in the collective memory across generations of Indians. The dots are said to represent the challenges of life and people have to weave their way out of the maze, in a harmonious manner as depicted by the synchronous Kolams. Figure 2 provides illustrations of Kolams.

Figure 2. Left: Kolams in front of the White House, Washington DC drawn by artist Shanthi Chandrashekar along with a group of organizers, volunteers, family, friends, and many communities from around the US, including the DC Public Schools. The Kolams consist of 1950 tiles, each of them being 12”x12” in size. Photo Credit- Tony Ventouris. Right: A 30”x22” Kolam by artist Shanthi Chandrashekar illustrated using pen and ink on paper. The illustrated Kolam is part of a series that was supported, in part, by a funding from the Montgomery County Government and the Arts and Humanities Council of Montgomery County, Maryland, USA. Photo Credit: Tony Ventouris. Kolams are sacred drawings and are a type of South Indian folk art. These are traditionally drawn on the floor everyday by women, using rice flour (such as the one shown in the figure on the left). They are drawn on the threshold of the house symbolically signifying the meeting of the external and internal positive energies. Kolams included after obtaining permission from artist Shanthi Chandrashekar.

This thought is further explored in the query about how Indian art is interconnected with the art of other cultures, where a respondent said that the art of most Indigenous cultures is connected in a basic way. The aim is to look for beauty, context, content, and core in all aspects of everyday life.

Thus, understanding the multi-dimensional nature of intelligence and incorporating cognitive aspects like empathy and emotions can be beneficial in the context of training ethical AI systems. Objective rules are useful in as much as they can serve as a baseline, but subjective elements like empathy and emotions have to be considered in order to enforce a sense of purpose, stability, and well being of all stakeholders involved in the AI pipeline. Further, understanding how humans traverse through everyday decision making (such as through past experiences, ancient knowledge repositories, formal education, etc.) can help enhance accountability in AI systems.

4.3. Discussion of RQ3

Can the Indian arts shed light on new forms of participatory approaches that can be useful in the AI pipeline?

In order to analyze this RQ, in our interview questionnaire, we had included questions concerning how artists collaborate with co-artists and accommodate their views to achieve harmony in their renditions, how they are able to reconcile contradicting opinions with their co-artists, how they cater to the requirements of diverse audience, and the extent of freedom they have in interpreting the art works of other artists. Below, we summarize the responses to these questions.

In response to the question concerning reconciliation of contradicting opinions (which often becomes necessary in settings involving stakeholders with differing requirements such as in an AI pipeline), one respondent cited the example of the proliferation of Harihara sculptures from the 5 century onwards and mentioned that often art was seen as a vehicle to transcend differences. The respondent wrote that the Harihara sculptures are indicative of the forces of fusion and syncretism in the Indian society. Ontologically, Harihara signifies that there is no metaphysical difference between the two principal divinities of Hinduism -Vishnu (Hari) and Shiva (Hara). Epistemologically, it represents the idea that diversity emerges from unity before merging itself into unity again. Another example of syncretism given by the respondent concerned a medieval sculpture at the Indian Museum in Kolkata which represents Harihara flanked by Buddha and Surya. In this, a single sculpture combines the four most important philosophies – Saiva, Bhagvata, Saura, and Saugata.

Adding to the aforementioned line of thought, another respondent quoted the Sanskrit phrase ‘spardhaaya vardhate vidya’, which roughly means competition enhances knowledge. Elaborating on this point, the respondent said that understanding the benefits of complementary viewpoints helps in convergence of divergent art forms, and in turn, helps in presenting a collaborative performance as one aesthetic entity.

In response to a question concerning how knowing about other art forms helps an artist to be more creative and accommodate other art forms, one respondent said that being aware of multiple art forms and immersing oneself in various art forms to the extent possible, enhances accommodation. A few other respondents mentioned that various art forms are intrinsically interconnected and an awareness of the same can help them in engaging creatively with their co-artists. Elaborating this point by taking the example of music based art forms, one respondent mentioned that Indian music reflects the metaphysical idea at the heart of Indian philosophy which is that humans are all fragmented parts of a whole, ultimately united as one. According to the respondent, this makes one approach music as a means to understand this oneness rather than merely as individual expressions of creativity. The respondent added that this attempt at connectivity can be seen in the Raga/tala system of Indian music, which is connected to nature and the environment, and ultimately to the essence of the universe. The respondent gave the example of ‘svaras’ in Indian music which refer to the musical notes on an octave, which are believed to have origins in the music of animals (notes: Re, Ga, Dha, Ni) and birds (notes Sa, Ma and Pa).

In response to a question concerning the extent of freedom artists have in interpreting works of other artists and incorporating the same in their renditions, one respondent mentioned that respect to the values of the original works is imperative. Another respondent said interpreting an art piece necessitates an analysis of the literal meaning and context of the piece, the cultural, philosophical and ideological leanings of the artist, the socio-political context, the layers of meaning in the art composition like the Padartha or the literal meaning, Vakyartha or the grammatical embeddings, Gudhartha or the hidden meaning, Visheshartha or the nuanced higher meaning, the symbolism, imagery, and the sub texts. The respondent mentioned that all these put together helps in understanding the totality of the art piece.

Based on the above responses, it can be inferred that recognizing and respecting subjective viewpoints is an essential requisite in any participatory framework. It is important to note that subjective viewpoints are inherently dynamic: they vary from person to person, place to place, and time to time. Thus, incorporation of subjective viewpoints can enhance the applicability of AI systems. Further, understanding the broader socio-cultural settings of a problem, understanding domain-specific requirements and challenges, and striving for harmony even in presence of differing stakeholder opinions can therefore be useful participatory paradigms to consider in the AI pipeline.

4.4. Discussion of RQ4

Can we learn new ways of designing adaptable and robust AI systems by looking at Indian artists and arts?

In order to analyze this RQ, we included a question that asked the artists to describe how they adapt to changing times, and to describe how change is adapted without disrupting the core of their art form.

In response to a question that asked how artists adapt to changing times, one respondent quoted one of India’s oldest poets Kalidasa who says ‘puranamityeva na sadhu sarvam, na chaapi kavyam navamitya vadhyam’

, a Sanskrit verse which means not everything ancient is always good and not everything modern is bad. The respondent mentioned that artists have twin responsibilities - to cater to the changing aesthetic sensibilities of the connoisseurs and secondly to mould the aesthetics of the audience and raise their standards of understanding the underlying values and beauty. This twin purpose should probably serve AI systems well too.

All artists have to adapt to changing times and even react to change. One respondent observed that the arts, particularly performing arts, cannot or should not freeze in time. The tradition becomes the unchanging core which gives the artist the capability to face changes. The respondent said that while modern tools and technologies can be helpful in furthering traditions and traditional art forms, the disruption can be caused only by the user, not the times or the changes it brings to the situation.

Changes in art are often brought about by changing tastes in society. The audience plays an important role in the Indian arts, with the artist’s role being to elevate the audience with their art to higher levels of emotional experience or even transcendence. One artist said that his teacher would tell him that while it is important to have an organic relationship with one’s audience, an artist should also independently set their own bar for excellence and artistic integrity. Using this as a metaphor in the context of how AI systems need to be trained, the respondent said that it is a worthwhile consideration to similarly encode independent ethical boundaries that are not influenced by training data from the real world. Quoting the example of face recognition algorithms that failed to work with dark skinned faces, the respondent said an ethical AI design should encode the need to recognize and support the diversity of faces on the planet and therefore recognize the lack of diversity in training data before it gets deployed.

Another respondent said the way in which art can both embrace and shock diverse audiences is an important metaphor for how AI could be designed. The respondent said a lot of AI technologies today facilitate the creation of bubbles - echo chambers of niche interests and ideologies, and added that AI must move away from identifying and grouping similar subsets, but should open new ways at looking at the world. The respondent said that an AI system should act like a friend in school who was a tastemaker, who was not interested in recommending music that the respondent liked, but instead would deliberately throw in something that one had never heard before and thereby open one’s mind to new experiences.

An oft quoted metaphor is that of the wheel, where the core at the center is fixed but the rim moves on. Most of the respondents subscribed to this notion of the continuity of the artistic tradition through changing times.

This characteristic of continuous adaptability without annihilation of diverse values is also needed in AI systems. Adaptability is thus a cumulative process that synchronously blends diverse principles in a gradual manner so as to accommodate change while maintaining continuity, which can be valuable in enhancing robustness of AI systems. Furthermore, expanding the horizons of AI system design by incorporating a broad world view, and exploration of new ethical paradigms could be seen as some ways of enhancing generalizability and robustness of AI systems.

5. Potential Pathways

In this section, we make several recommendations for the design, development, and evaluation of ethical AI systems based on our insights gained through the study.

5.1. Integrating multimodal data for characterizing ethics

Several sources are leveraged for the process of knowledge transfer in traditional settings: study of ancient texts, oral lessons, practice sessions, and demonstrations, to name a few. We posit that in order to develop ethical AI systems, a similar multimodal approach is necessary. This not only entails providing multimodal explanations underlying an AI decision and training systems through demonstrations, but also leveraging multimodal data sources (text, image, video, audio) encompassing diverse ethical perspectives in defining, designing, and developing ethical AI systems. Existing works only address parts of the proposed strategy such as through multimodal explanations (Park et al., 2018; tsushi Kanehira et al., 2019), exploration of fairness related issues in multimodal settings (Chen et al., 2020; Wang et al., 2021), inference of norms from stories (Frazier et al., 2020)

, or by using principles of imitation learning towards value alignment

(Taylor et al., 2016) ; these works neither leverage multimodal data sources nor incorporate diverse ethical perspectives in their design. For example, oral traditions of teaching in Indian art education are heavily informed by “Dhvani” or sound signals. Yet, audio signals are seldom leveraged in the design of ethical AI systems. Studies in psychology have emphasized the role of sound in problems related to causal attributions (Gerstenberg, 2021), which are very pertinent in questions related to accountability. Recent studies have highlighted the gaps in applying existing algorithmic fairness approaches to real world applications (Bakalar et al., 2021)

. Thus, it becomes necessary to go beyond the current emphasis on a selected subset of data modalities in defining ethical notions (such as fairness based on statistical parity or equalized odds using categorical data) and leverage diverse data modalities such as audio and images as well in such characterizations whenever possible. More broadly, we believe that leveraging multimodal data during all stages of the AI pipeline—from problem formulation and data curation to algorithm design and evaluation–can be beneficial in capturing and reflecting diverse ethical values, which in turn can help in narrowing the gap between research and practice.

5.2. Consistent life-long learning

Given their similarity to teacher-taught techniques, we premise that knowledge distillation techniques and student-teacher learning can be useful analogies in the design of ethical AI systems (Wang and Yoon, 2021). If one can think of the entire body of knowledge concerning ethics in the world as a large teacher model, then the goal is to incorporate the values of this large teacher model into a small student model, which in this context is an ethical AI system, as accurately as possible. Of course, this computational abstraction has limitations–it cannot possibly capture all the nuances associated with the complex body of world knowledge pertaining to ethics– but nevertheless can be a starting point given its semblance to the teacher-taught techniques of art education. It is important to notice that the learning process in traditional art education spans many years, sometimes even decades. Thus, in order to be able to effectively leverage the benefits of such teacher-taught systems, AI models need to be trained over long regimes, possibly even requiring life-long learning (Parisi et al., 2019; Lange et al., February 2021)

. A recent study supports this claim by demonstrating that “patience” (training over large epochs) and “consistency” (teacher and student should process the exact same data and match on a large number of support points to generalize well) are important in improving the effectiveness of knowledge distillation techniques in practical settings

(Beyer et al., 2021). Training over long regimes and ensuring consistency between the ethical models/theories out in the world and their computational abstractions will thereby help in enhancing AI ethics.

5.3. Incorporating Holistic Viewpoints

While data-driven techniques operate bottom up, recently neuro-symbolic techniques (Rocktaschel et al., 2015), (Sikka et al., 2020) have been developed that impose top-down constraints on AI models. These methods use first-order logic to regularize the bottom up learning so that it conforms to background knowledge such as ‘a human can ride a horse’, and ‘a horse cannot ride a human’. Similarly causal models (Pearl, 2009)

have also been proposed which also operate on the principle of using common sense or background knowledge to guide the bottom up learning. Such techniques offer the promise of incorporating ethics top-down into the AI pipeline, since ethics is also a form of background knowledge. The advantage of neuro-symbolic techniques is that they retain the accuracy of the deep learning frameworks while injecting background knowledge. If ethics can be expressed in term of first order logic or simple causal frameworks, then very likely such methods will enable organic incorporation of ethics into AI. The challenge is of course that not all ethical frameworks can be captured in first order logic. This challenge becomes especially acute in the case of Indian art in which underlying principles are either stated at a high semantic level that are beyond current knowledge representations or are simply understood in an implicit fashion from multiple instantiations. The challenge for AI systems going forward to is to incorporate such a holistic view of ethics into the learning framework so that ethical considerations are baked into the system.

5.4. Facilitating Adaptation without Annihilation

The lifelong learning described above also has the dimension of the dynamism of the art form. Indian art is the result of constant dialog between the creator and the audience, and therefore has evolved as the tastes of the audience have evolved. At the same time, it has preserved its essential nature, so it is still recognizable as Indian art. In other words, it is a work in progress that maintains a constant identity while also constantly changing. Part of such fluidity also stems from the integral role of improvisation in Indian art. No two Indian musical performances even by the same artist are the same since the artist improvises while adhering to overarching principles. Ethics offers a similar challenge of maintaining a recognizable core while adapting to changing times and unanticipated needs. AI ethics is a shared, dynamic, cumulative process as opposed to just being a static framework. AI systems need to go beyond static training sets to dynamic learning in which the model is able to adapt to the changes in the data over time while maintaining adherence to ethical principles. Some of this is seen in modern system that seek to prevent catastrophic forgetting (Raghavan et al., 2020), however ethics offers a much richer semantic framework than such systems can accommodate. That said, they do offer a methodological avenue forward in that they point the way to a more expansive definition of both the tasks underlying ethics and the notion of catastrophic forgetting and maintaining adherence to essential principles.

5.5. Identifying Commonalities

Note that there is much in common between ethics inspired by Indian art and ethics from the Western canon. The bottom up nature of data driven learning in fact provides a bridge between explicit ethical principles and principles stemming from practice. Explicit ethical principles are also grounded in experience and can be thought of as resulting from human distillation of experience. Harmonization of ethics stemming from immediate experience and ethics stemming from years or decades of experience requires careful understanding of what is in common and what is not between the two. Identifying commonalities will involve all the research challenges described above, and is important because it will enable a knowing incorporation of ethics into AI systems. Identifying commonalities across cultures and infusing the same into AI system design will also enhance the applicability of such systems across geographies. Indeed, the Indian principle of ‘Unity in Diversity’ which has been instrumental in maintaining the continuity of diverse Indian art traditions (despite the tests of time), serves as an illustration to this point.

6. Limitations

The study reflects the background of the authors and the interview respondents. The study was carried out with an intention of identifying how the overarching principles of Indian arts can inform the FAccT community. We thus do not delve deep into any given Indian art form such as music or dance, and do not necessarily provide a comprehensive coverage of either Indian arts or its possible influence on AI ethics. A wider survey of Indian art might reveal aspects we have left uncovered. The technical aspects addressed here are thus only a subset of the larger requirements of ethical AI systems. An analysis of specific works of Indian art could potentially offer a rich avenue for further research. The study focused on one non-Western art form, inclusion of other non-Western art forms (beyond Indian arts) can shed light on other valuable abstractions for ethical AI system design and development.

7. Conclusions

Exclusion of non-Western perspectives in ethical AI system design and development creates an imbalance in that it not only sidelines the values and principles of these communities, but also fails to understand the broader historical, cultural, and social narratives relevant to a problem, thereby resulting in adverse consequences. In this regard, the field of arts has the potential to shed light on diverse and unique socio-cultural-historical narratives, thereby serving as a bridge across research communities. In this work, through qualitative interviews of artists, art scholars, and researchers of diverse Indian art forms such as music, dance, sculpture, painting, floor drawings, etc., we analyzed how non-Western ethical abstractions, methods of learning, and participatory approaches observed in Indian arts, one of the most ancient yet perpetual and influential art traditions, can inform the FAccT community. Insights from our study outline the need for (1) incorporating holistic approaches in ethical AI algorithms, (2) integrating multimodal data formats for ethical AI system design and development, (3) facilitating adapatability without annihilation of ethical values, (4) consistent life long learning to enhance AI accountability, and (5) identifying commonalities across cultures and times to enhance generalizability of ethical AI systems.

We thank Anupama Kailash, Arun Subramaniam, Cristian De Vietri, Krish Ashok, Manish Maheshwar, Dr. Pappu Venugopala Rao, Pieter Weltevrede, Dr. Sai Ramakrishna Susarla, Shanti Chandrasekar, Shubhendra Rao, Dr. Subhash Kak, Dr Sundar Sarukkai, Suresh Vyas, Vijaya Krish, Dr. Vinod Vidwans, and anonymous respondents who participated in the interview study.


  • R. Abebe, K. Aruleba, A. Birhane, S. Kingsley, G. Obaido, S. L. Remy, and S. Sadagopan (2021) Narratives and counternarratives on data sharing in africa. FAccT. Cited by: §1.
  • C. Bakalar, R. Barreto, S. Bergman, M. Bogen, B. Chern, S. Corbett-Davies, M. Hall, I. Kloumann, M. Lam, J. Q. Candela, M. Raghavan, J. Simons, J. Tannen, E. Tong, K. Vredenburgh, and J. Zhao (2021) Fairness on the ground: applying algorithmic fairness approaches to production systems. ArXiV. Cited by: §5.1.
  • L. Beyer, X. Zhai, A. Royer, L. Markeeva, R. Anil, and A. Kolesnikov (2021) Knowledge distillation: a good teacher is patient and consistent. ArXiv. Cited by: §5.2.
  • E. Bietti (2020) From ethics washing to ethics bashing: a view on tech ethics from within moral philosophy. ACM FAccT. Cited by: §1.
  • R. Binns (2018) Fairness in machine learning: lessons from political philosophy. ACM FAccT. Cited by: §1.
  • C. Boyes-Watson (2014) Center for restorative justice. Suffolk University, College of Arts and Sciences. Cited by: §4.1.
  • J. Chen, I. Berlot-Attwell, X. Wang, S. Hossain, and F. Rudzicz (2020) Exploring text specific and blackbox fairness algorithms in multimodal clinical nlp. EMNLP. Cited by: §5.1.
  • P. Cihon, M. M. Maas, and L. Kemp (2020) Should artificial intelligence governance be centralised? design lessons from history. AIES. Cited by: §1.
  • A. Coomaraswamy (1975) The aims of indian art. Studies in Comparative Religion. Cited by: §2.
  • K. Crawford and T. Paglen (2019) Excavating AI: The Politics of Images in Machine Learning Training Sets. Cited by: §1.
  • S. croeser and P. Eckersley (2019) Theories of parenting and their application to artificial intelligence. AIES. Cited by: §1.
  • A. Daniele and Y. Song (2019) AI+art= human. AAAI AI Ethics and Society. Cited by: §1.
  • P. P. Dhar (2011) Indian art history—changing perspectives. D.K.Printworld (P) Ltd. and National Museum Institute, New Delhi. Cited by: §2.
  • P. Francastel (2000) Art and technology in the nineteenth and twentieth centuries. New York, Zone. Cited by: §1.
  • S. Frazier, M. S. A. Nahian, M. O. Riedl, and B. Harrison (2020) Learning norms through stories: a prior for value aligned agents. AIES. Cited by: §5.1.
  • T. Gerstenberg (2021) Going beyond the here and now: counterfactual simulation in human cognition. Cited by: §5.1.
  • T. Gorichanaz (2020) Engaging with public art: an exploration of the design space. CHI. Cited by: §1.
  • J. Graham, J. Haidt, S. Koleva, M. Motyl, R. Iyer, S. P. Wojcik, and P. H. Ditto (2013) Moral foundations theory: the pragmatic validity of moral pluralism. Advances in experimental social psychology. Cited by: §4.1.
  • L. M. Hampton (2021) Black feminist musings on algorithmic oppression. ACM FAccT. Cited by: §1.
  • L. Hancox-Li and I. E. Kumar (2021) Epistemic values in feature importance methods: lessons from feminist epistemology. ACM FAccT. Cited by: §1.
  • A. Hanna, E. Denton, A. Smart, and J. Smith-Loud (2020) Fairness and machine learning. FAccT. Cited by: §1.
  • K. M. Higgins (2008) An alchemy of emotion: rasa and aesthetic breakthroughs. The Journal of Aesthetics and Art Criticism. Cited by: §2.
  • D. D. Hudson (2008) The body of god: an emperor’s palace for krishna in eighth-century kanchipuram. Oxford University Press. Cited by: §2.
  • S. L. Huntington and J. C. Huntington (2016) The art of ancient india - buddhist, hindu, jain. Motilal Banarsidass Publishers Private Limited, New Delhi. Cited by: §2.
  • S. Inayatullah (2006) Culture and fairness: the idea of civilization fairness. Fairness, Globalization and Public Institutions. University of Hawaii Press. Cited by: §4.1.
  • A. Jacobs and H. Wallach (2021) Measurement and fairness. ACM FAccT. Cited by: §1.
  • B. Jensen (2020) How ai and art hold each other accountable. Stanford HAI Blog. Cited by: §1.
  • B. Jensen (Retrieved 2021) HAI Visiting Artist Rashaad Newsome: Designing AI with Agency. Cited by: §1.
  • E. S. Jo and T. Gebru (2020) Lessons from archives: strategies for collecting sociocultural data in machine learning. ACM FAccT. Cited by: §1.
  • S. Kak (2021) The śiva sūtra: play of consciousness. Brahmavidya: The Adyar Library Bulletin. Cited by: §2.
  • T. Krafft, M. Hauer, L. Fetic, A. Kaminski, M. Puntschuh, P. Otto, C. Hubig, T. Fleischer, P. Grünke, R. Hillerbrand, C. Hustedt, and S. Hallensleben (2020) From principles to practice - an interdisciplinary framework to operationalise ai ethics. pp. . Cited by: §1.
  • M. D. Lange, R. Aljundi, M. Masana, S. Parisot, X. Jia, A. Leonardis, G. Slabaugh, and T. Tuytelaars (February 2021) A continual learning survey: defying forgetting in classification tasks. IEEE Transactions on Pattern Analysis and Machine Intelligence. Cited by: §5.2.
  • S. Leavy, E. Siapera, and B. O’Sullivan (2021) Ethical data curation for ai: an approach based on feminist epistemology and critical theories of race. AIES. Cited by: §1.
  • G. Leech (2011) James Bridle – Waving at the Machines. Cited by: §1.
  • A. Leroi-Gourhan (1993) Gesture and speech. MIT Press. Cited by: §1.
  • S. Maira (2006) Towards ananda-rethinking indian art and aesthetics. Viking Publishers. Cited by: §2, §2.
  • A. Meshi (2021) Deconstructing Whiteness.

    CVPR Workshop on Ethical Considerations in Creative Applications of Computer Vision

    Cited by: §1.
  • S. Miller - Carpenter, A. Cepak, and Z. Peng (2018) An exploration of the complexity of journalistic interviewing competencies. Journalism Studies 19, pp. 2283–2303. External Links: Document Cited by: §3.
  • D. Novitz (1987) Knowledge, fiction, and imagination. Temple University Press. Cited by: §1.
  • G. I. Parisi, R. Kemker, J. L. Part, C. Kanan, and S. Wermter (2019)

    Continual lifelong learning with neural networks: a review

    Neural Networks. Cited by: §5.2.
  • D. H. Park, L. A. Hendricks, Z. Akata, A. Rohrbach, B. Schiele, T. Darrell, and M. Rohrbach (2018) Multimodal explanations: justifying decisions and pointing to the evidence. CVPR. Cited by: §5.1.
  • M. Patel, H. Webb, M. Jirotka, A. Davoust, R. Gales, M. Rovatsos, and A. Koene (2019) Harnessing interdisciplinarity to promote the ethical design of ai systems. European Conference on the Impact of Artificial Intelligence and Robotics - Oxford, United Kingdom. Cited by: §1.
  • S. Paul (2015) The fallacy of the fairness concept. The Hindu. Cited by: §4.1.
  • J. Pearl (2009) Causality: models, reasoning and inference, 2nd edition. Cambridge University Press. Cited by: §5.3.
  • A. Raghavan, J. Hostetler, I. Sur, A. Rahman, and A. Divakaran (2020) Lifelong learning using eigentasks:task separation, skill acquisition, and selective transfer. 4th Lifelong Machine Learning Workshop, Proceedings of the 37th International Conference on Machine Learning (ICML), PMLR, 8.. Cited by: §5.4.
  • A. Rajgopalan (2021) Ancient indian education: the para-apara vidya continuum. International Journal of Applied Research. Cited by: §2.
  • I. D. Raji, M. K. Scheuerman, and R. Amironesei (2021) You can’t sit with us: exclusionary pedagogy in ai ethics education. FAccT ’21, New York, NY, USA, pp. 515–525. External Links: ISBN 9781450383097, Link, Document Cited by: §1.
  • Rocktaschel, Singh, and Riedel (2015) Injecting logical back-ground knowledge into embeddings for relation extraction.. ICML. Cited by: §5.3.
  • N. R. A. Romm (1998) Interdisciplinary practice as reflexivity. Systemic Practice and Action Research 11 (1), pp. 63–77. Cited by: §1.
  • N. Sambasivan, E. Arnesen, B. Hutchinson, T. Doshi, and V. Prabhakaran (2021) Re-imagining algorithmic fairness in india and beyond. ACM FAccT. Cited by: §1, §4.1.
  • R. Schechner (2001) Rasaesthetics. The Drama Review. Cited by: §2.
  • S. Schwartz (2004) Rasa: performing the divine in india. New York: Columbia Unversity Press. Cited by: §2.
  • E. Shaken (2002) Art in the information age: technology and conceptual art.. Leornado. Cited by: §1.
  • K. Sikka, A. Silberfarb, J. Byrnes, I. Sur, E. Chow, A. Divakaran, and R. Rohwer (2020) Deep adaptive semantic logic (dasl) : compiling declarative knowledge into deep neural networks. Cited by: §5.3.
  • A. P. Singh (2012) Indian perspectives on intelligence: some psychological and philosophical evidences. International Conference on Cognitive Science. Cited by: §4.2.
  • R. Srinivasan, E. Denton, J. Famularo, N. Rostamzadeh, F. Diaz, and B. Coleman (2021) Artsheets for art datasets. NeurIPS Datasets and Benchmarks Track. Cited by: §1.
  • R. Srinivasan and K. Uchino (2021) The role of arts in shaping ai ethics. AAAI Workshop on Reframing Diversity in AI: Representation, Inclusion, and Power. Cited by: §1.
  • L. Stark and K. Crawford (2019) The work of art in the age of artificial article intelligence:what artists can teach us about the ethics of data practice. Surveillance and Society. Cited by: §1.
  • O. Svyetlov (2001) Interview as a method and its application in journalism. Cited by: §3.
  • J. Tasioulas (2021) The role of the arts and humanities in thinking about artificial intelligence (AI). Cited by: §1.
  • J. Taylor, E. Yudkowsky, P. LaVictoire, and A. Critch (2016) Alignment for advanced machine learning systems. Machine Intelligence Research Institute. Cited by: §5.1.
  • The-Art-Story (Retrieved November 2021) Post modern art— modern art terms and concepts. Cited by: §2.
  • tsushi Kanehira, K. Takemoto, S. Inayoshi, and T. Harada (2019) Multimodal explanations by predicting counterfactuality in videos. CVPR. Cited by: §5.1.
  • K. Vatsyayan (1997) The square and circle of the indian arts. Roli Books International. Cited by: §2.
  • K. Vatsyayan (Retrieved 2021) Inter-relationship of the arts. Cited by: §2, §2.
  • N. M. K. Venkat (2010) Lyrical musings on indic culture, a sociological study of songs of sant purandara dasa. Readworthy Publications Pvt Ltd. Cited by: §4.1.
  • S. Venkatasubramanian and M. Alfano (2020) The philosophical basis of algorithmic recourse. ACM FAccT. Cited by: §1.
  • P. Venugopal (2018) Research methodology for music, dance, theater. University of Silicon Andhra. Cited by: §1, §3.
  • J. Wang, Y. Liu, and X. E. Wang (2021) Assessing multilingual fairness in pre-trained multimodal representations. ArXiV. Cited by: §5.1.
  • L. Wang and K. Yoon (2021) Knowledge distillation and student-teacher learning for visual intelligence: a review and new outlooks. IEEE PAMI. Cited by: §5.2.
  • J. O. Young (2001) Art and knowledge. Routledge, London, UK. Cited by: §1.