Out of the Black Box: Indigenous Protocols for AI | Abstract
IP//AI: ABSTRACT
In this paper we share our journey starting with an international group of Indigenous technologists at the inaugural workshop series in Hawaii in 2019, leading to the IP//AI Incubator in March, 2021.
Key learnings from the foundations of these works were the need for Indigenous AI to be regional in nature, conception, design and development, to be tethered to localised Indigenous laws inherent to Country, to be guided by local protocols to create the diverse standards and programming logic required for the developmental processes of AI, and to be designed with our future cultural interrelationships and interactions with AIs in mind.
Through Country Centered Design we established some broad principles and protocols and then moved towards a test case, running some preliminary trials applying an Aboriginal kinship system as a selection framework in genetic computing. Our findings throughout this process were encouraging, indicating that there is potential for Indigenous Knowledge to guide the design and engineering principles and practices of AI, bridging the current ontological and epistemological divides between machines, humans and the environment...
Extended paper via link here
Professor Angie Abdilla is a palawa~trawlwoolway designer/technologist, CEO of Old Ways, New and is a member of the Global Future Council on Artificial Intelligence for Humanity for the World Economic Forum.
Megan Kelleher belongs to the Baradha and Gabalbara peoples is a PhD Candidate investigating blockchain and Indigenous governance.
Rick Shaw is a Gamillaroi mathematician.
Dr Tyson Yunkaporta is a boy who belongs to the Apalech clan from Far North Qld, author, Indigenous researcher and the founder of the Indigenous Knowledge Systems Lab, NIKERI, Deakin University.
IP//AI: BACKGROUND
The Indigenous Protocols and Artificial Intelligence (IP//AI) Incubator - spawned out of the international cohort and work established in 2018 (Lewis et al, 2020) - brought together a discrete group of Australian Indigenous peoples to further some of the key insights garnered from this prior work. One of the key insights identified was the need to progress toward the materialisation of a practical and tangible format that reflected Indigenous peoples' future dreaming of what AI could become, presented to the broader global AI community. Another key insight revealed the need for the work to be grounded by our relational connections to diverse territories as Indigenous peoples worldwide and based upon these discrete cultural identities.
The third iteration of the IP//AI work took the form of an incubator. We invited a diverse group of Australian Aboriginal researchers, professionals and practitioners to commune virtually online as we facilitated a regional approach to the work. We specifically prefaced localised cultural knowledges, systems and protocols, perspectives, environmental needs and social conditions relating to this continent, intentionally seeking to reveal what an Australian Aboriginal AI could become. We aimed to achieve this through experimentation and prototyping how our protocols could deliver alternatives to the usual desired outcomes of automation…
Extended paper via link here
IP//AI: INTRODUCTION
Indigenous peoples' connection to the physical, spiritual and sentient worlds are based upon ontologically and epistemologically divergent frameworks, including sensing and presencing. These complex, relational connections to Country and kinship networks simultaneously align us, while also shaping our discrete cultural identities through Indigenous laws, languages and protocols determined by the nature of Country itself. As Indigenous peoples, we make sense of the world and act as its custodians by following the Law of the land. This guides our lives and work, not only when we’re out on the rivers and plains, but also when we are working online to create an approach to the conceptual design and software engineering principles within AI. Through Indigenous governance, standards and protocols we hope to contribute to the evolution of technology, its philosophy and engineering methodologies by prioritising and centring Country within automated systems and machines. Through the linkage of Indigenous techno philosophies to sector standards and best practices, a more equitable and healthy relationship between Country, humans and technology may be possible.
From an Indigenous worldview that privileges communal wellbeing, wholeness and balance, we explored Western cultural notions of ‘intelligence’ within AI to begin creating an alternate conceptual foundation - principles and processes that support our future dreamings of AI. This foundation is informed by what we call our ‘old ways’, or Traditional knowledge systems, in which technology design precedents embody relational connections between Country and kin…
Extended paper via link here
IP//AI: GLOSSARY
It is necessary to include a glossary for two reasons. Firstly, for clarity of terminology - not because the terms used are unfamiliar, but because they mean different things within different cultural contexts and disciplines. For example, what ‘protocol’ means to Indigenous people is likely to be entirely different within a computer science environment (eg. refer Australian Human Rights Commission 2021, p. 17; and Mattingly-Jordan et al, n.d.). Therefore, this glossary serves a function more akin to disambiguation than establishing a single semantic interpretation of terms.
In so doing, this glossary performs its second function within this text: demonstrating the instantiation of two coded ontologies - one ethnocultural and another digital, and the need for a process to intermediate and align these codes. It is also honouring an Indigenous Australian principle: a common protocol when groups with different codes meet is that the participants will establish their languages and agree on the standards that will be in place during the ritual of exchange…
Extended paper via link here
IP//AI: METHODOLOGY
Country Centred Design (CCD) was developed as an alternative to the human-centred design processes which preface the needs of humans above and beyond any other living being, element, entity and or natural system (Old Ways, New, 2016). The Indigenous-led process comprises four key cycles: culture, research, strategy and technology, reflecting the nature of our relationship with natural, complex systems. The CCD methodology has been tested in a variety of contexts and iteratively developed to be structurally flexible for utility while providing specificity as required. Indigenous knowledge systems, cultural practices and design principles guide the cyclic methodology, all the time prioritising and centering the needs of Country and respecting its agency and autonomy as an intelligent entity.
The first cycle, culture, begins by developing relationships with Traditional Custodians and their communities; utilising strategic stakeholder engagement practices which assist in building a cultural understanding of Country from deep time, colonial, post-colonial, present and future dreamings at the heart of the project. Adherence to the cultural protocol of beginning with appropriate Indigenous Custodian consultation brings the group under customary authority and guidance to ensure cultural integrity in project design.
For the IP//AI incubator project, the initial series of virtual workshops unfolded under Elder authority on Gadigal/Wangal lands (Sydney, Australia). Aunty Bronwyn Penrith (Wiradjuri/Yuin and Gadigal bloodlines) stewarded the work….
Extended paper via link here
P//AI: INDIGENOUS DESIGN PRINCIPLES FOR DEEP LEARNING
In the story of our collaborative deep learning experience (in the Indigenous sense of deep learning as inhabiting the ontologies of human and non-human entities within diverse systems) there is a preliminary experiment in genetic computing that attempts to apply Indigenous methodologies to ‘breeding’ algorithms. Genetic computing seeks to leverage evolutionary processes in AI systems. Because of limitations we perceived in all axiomatic systems, we regarded this as a quixotic quest, as we did not believe that any mechanically built system could achieve the state of complexity that produces emergent qualities.
Our Indigenous perspective on this problem was that a system can only be considered in its entirety, and that technology can never be more than the sum of its parts. Further, the ‘meaning’ of a system is not separate from the system itself, which we saw as a limitation in modelling exercises. Always grounding our thoughts in Country, we aligned these concerns with Mandelbrot’s observation (Milnor, 1989) that for a map of an area to contain all the features of that area, the map would have to be as big as the area itself. We agreed that the quest by some in physics for a “Theory of Everything” (Hawking, 2006), a coherent set of rules that can describe all phenomena, is not achievable as a thing separate from the whole universe. In Country Centred Design, you can never stand outside a system and observe or intervene - you must embrace the fact that you are part of that system…
Extended paper via link here
IP//AI: STORY AND FINDINGS OF THE IP//AI INCUBATOR
Protocols in Indigenous cultures are not simply commandments or statements of norms such as, “Look after Country and Country will look after you”. Protocols are quite specific and deal with relational behaviours that must be adhered to in specific contexts (e.g. Where does a young person direct their voice and gaze when in the presence of an opposite gender Elder of high status?). In our yarns to begin outlining and testing various protocols, we commenced quite broadly with general Indigenous ethics, building on previous work in this area (Lewis et al, 2020) which was a good ethical starting point but did not elucidate the specificity that software engineers require to inform programming standards, protocols or, rules as code.
We began this work with an awareness of Indigenous data sovereignty issues an emergent field involving Indigenous control over the protection and use of data that is collected from our communities, including statistics, cultural knowledge and even user data. We nicknamed our desired protocols around this as ‘Blackfella box’, referencing the idea of ‘black boxes’ in digital systems; this was our shorthand for identifying areas of knowledge that we could not share with the world. We also flagged this as a potential test case for automating a restricted knowledge protocol, possibly using a blockchain/smart contract application that we referred to as “Proof of Aunty”, asserting that proof of cultural authority is more of a priority for us than proof of stake or proof of work....
Extended paper via link here
IP//AI: INSIGHT> INDIGENOUS ETHICS> STANDARDS> PROTOCOLS> CODE
While defining the intention, purpose and creation story of tools driving autonomous decision-making utilities, developmental and cultural protocols began to merge. Once we worked through each AI subset or domain and its discrete processes, languages and related syntax for code we were able to translate the ethics to specific standards and developmental protocols. This enabled the translation of protocols for rules and code to evolve.
The avoidance protocol application was flagged as interesting but was not carried forward in the next iteration towards a prototype. The next developmental phase was informed by the moieties and sub-moieties of the kinship system. We applied these to a thought experiment we called ‘Sugarbagscape’ in which we reimagined the original agent-based modelling software Sugarscape (Epstein & Axtell, 1996) as being designed with kinship protocols for reproduction and land/resource management.
The agent-based modelling software was envisioned as having kinship/totemic/clan territories embedded in a landscape as law (protocols) to regulate the lawful behaviour of agents. Agents would have a limited set of protocols governing metabolic cycles, breeding cycles, peak harvest cycles, migration cycles etc. adjusted to different settings. Human agents would have ritual cycles added, as well as cycles for burning/regenerating particular ecosystems in the landscape. Plants would have the migration cycle set to zero…
Extended paper via link here
