1 of 29

strategies of listening

Debashis Sinha

deb@debsinha.com / sinhad@torontomu.ca

“hanuman flies over the himalayas”

generated using CLIP

2 of 29

a frame

3 of 29

in conversation with sound artist Budhaditya Chattopadhyay in

“Sound Practices in the Global South: Co-listening to Resounding Plurilogues”

“I have in mind a picture of a body in stillness and attention, breath slowed, brain function razor sharp and focused, ears open, allowing the periphery of our attention to be accepted and integrated into the act of listening.”

4 of 29

what does it mean to listen in this algorithmic space?

5 of 29

Ouzounian, G. 2020. Counterlistening. English Studies in Canada 46(2–4): 311–17. https://doi.org/10.1353/esc.2020.a903549.

(counterlistening)

Counterlistening provides a methodology for working with the output of algorithmic tools - a reminder of the strategy to actively seeks out what's been concealed, ignored, or denied - opening pathways to alternative possibilities, material, and knowledge

Counterlistening (Gascia Ouzounian): A "subversive, dangerous, and self-sacrificing" practice that seeks to hear what "should" remain unheard - what is ignored, concealed, or denied.

Application to AI Systems

Speculative and futuristic listening for possible worlds in the margins
Reveals hidden aspects of neural networks and machine learning processes
Makes manifest alternative possibilities that algorithms obscure

Counterlistening mirrors creativity:

Uncovers what's not immediately apparent
Creates "lightning bolt" discoveries that reframe understanding
Resists "fixity and stability of analysis"
Perfectly intertwined with creative process

6 of 29

Robinson, D. 2020. Hungry Listening: Resonant Theory for Indigenous Sound Studies. Minneapolis, MN: University of Minnesota Press.

(sovereign listening)

Sovereign Listening - A framework from Dylan Robinson's work that challenges the demand for cultural objects to present themselves in forms that accommodate "settler recognition"

The Problem

Traditional listening and knowledge approaches impose universality and demand that all knowledge be accessible to everyone
This extends to machine learning tools that require objects to be "fully apprehensible" through dominant frameworks
Creates tension for creators whose work may only be fully understood through culturally specific experience

Practical Applications

For Working with Algorithmic Tools:

Recognize the limits of quantized knowledge systems
Use algorithmic outputs as starting points, not endpoints
Integrate alternative ways of knowing based on authenticity, vulnerability, and community
Accept that some forms of knowledge lie outside algorithmic frameworks

7 of 29

Robinson, D. 2020. Hungry Listening: Resonant Theory for Indigenous Sound Studies. Minneapolis, MN: University of Minnesota Press.

(listening as territorial entry)

The act of listening as entering a territory as a guest rather than approaching sound as an object to be fully understood or possessed can guide our navigation in the AI space.

Listening as Territorial Entry: When you listen, you're not consuming or analyzing an object - you're entering into a space that belongs to someone else, with its own rules, histories, and ways of being.

Guest Status: As a listener, you are always a visitor in this sonic space. This guest status comes with both limitations and responsibilities - you cannot expect to understand everything, nor should you demand that the territory adapt to your ways of knowing.

We should approach AI outputs not as neutral territories we can fully map and understand, but as spaces created through specific cultural and technical processes.

8 of 29

“…to listen is to adopt an attitude of decoding what is obscure, blurred or mute, in order to make available the underside of meaning.”

- Roland Barthes in collaboration with Roland Havas, “Listening”, 1976

9 of 29

Our urgency arises from humans’ capacity to use AI as a knowledge system to create irrefutable “algorithmic truths” to reinforce domination. In doing so, other systems of knowledge production and other visions are denied and erased, as are other peoples’ agency, autonomy, and contestation. In this way AI coloniality extends beyond data colonialism…In insisting on a decolonial AI, we stand for the right of each historically marginalized community to reshape reality on their terms.

AI Decolonial Manyfesto

10 of 29

(what happens if)

11 of 29

(what happens if I interrupt the training)

12 of 29

(what happens if I present the model something it hasn’t been trained on)

13 of 29

(what happens if I alter the dataset, or work with small ones)

14 of 29

(what happens if I change the number within these brackets / or those)

15 of 29

(what happens if I create a feedback loop)

16 of 29

(what happens if I take the checkpoints and re-train the model on those)

17 of 29

(what happens if I turn off the click track / when I turn it on)

18 of 29

(what happens if I invert the MIDI data)

19 of 29

(what happens if I use my left hand instead of my right to work the mouse)

20 of 29

(what happens if I choose this gong, or that one)

21 of 29

(what happens if I turn this drum upside down)

22 of 29

(what happens if I consider the minerals my computer is made out of)

23 of 29

(what happens if I consider the wages the workers are paid)

24 of 29

(what happens if I consider a different story)

25 of 29

(what happens if I ask something of the spirit of my grandmother)

26 of 29

(what happens if I listen to the land)

27 of 29

(what happens if I listen to the air)

28 of 29

(what happens if I listen to my heart)

29 of 29

current/upcoming @ debsinha.com

deb@debsinha.com

sinhad@torontomu.ca

“Yama and Minos judge the dead, observed by the gods of the Hindu pantheon”

epoch 10, generated using CLIP

cover art “Adeva_v000_04” (Establishment, 2021)