Abstract

Today, in the post-industrial society, the data created by computers and the internet are becoming a new ‘means of production’. With this, there is a new- found realisation on the question of its ownership, distribution and protection. The question of accessibility in this regard is also becoming a crucial issue in democratic political systems. Internet and data have led to epistemological changes such that new rights for citizens are required with an overhaul of democratic processes founded in digital methods. It will definitely have its own challenges, limitations and possibilities. As data protection bills are now being introduced in many countries of the world, it is important to analyse the key theoretical nuances in the data ownership discourse. In this regard, this article attempts to highlight these nuances along with the scope for new public policy frameworks required in the domain of data ownership, distribution and protection.

Abstract

This paper considers the contextual multi-armed bandit (CMAB) problem with fairness and privacy guarantees in a federated environment. We consider merit-based exposure as the desired fair outcome, which provides exposure to each action in proportion to the reward associated. We model the algorithm's effectiveness using fairness regret, which captures the difference between fair optimal policy and the policy output by the algorithm. Applying fair CMAB algorithm to each agent individually leads to fairness regret linear in the number of agents. We propose that collaborative -- federated learning can be more effective and provide the algorithm Fed-FairX-LinUCB that also ensures differential privacy. The primary challenge in extending the existing privacy framework is designing the communication protocol for communicating required information across agents. A naive protocol can either lead to weaker privacy guarantees or higher regret. We design a novel communication protocol that allows for (i) Sub-linear theoretical bounds on fairness regret for Fed-FairX-LinUCB and comparable bounds for the private counterpart, Priv-FairX-LinUCB (relative to single-agent learning), (ii) Effective use of privacy budget in Priv-FairX-LinUCB. We demonstrate the efficacy of our proposed algorithm with extensive simulations-based experiments. We show that both Fed-FairX-LinUCB and Priv-FairX-LinUCB achieve near-optimal fairness regret.

Abstract

Restless multi-armed bandits (RMABs) generalize the multi-armed bandits where each arm exhibits Markovian behavior and transitions according to their transition dynamics. Solutions to RMAB exist for both offline and online cases. However, they do not consider the distribution of pulls among the arms. Studies have shown that optimal policies lead to unfairness, where some arms are not exposed enough. Existing works in fairness in RMABs focus heavily on the offline case, which diminishes their application in real-world scenarios where the environment is largely unknown. In the online scenario, we propose the first fair RMAB framework, where each arm receives pulls in proportion to its merit. We define the merit of an arm as a function of its stationary reward distribution. We prove that our algorithm achieves sublinear fairness regret in the single pull case , with being the total number of episodes. Empirically, we show that our algorithm performs well in the multi-pull scenario as well.

Abstract

Existing approaches to fairness in stochastic multi-armed bandits (MAB) primarily focus on exposure guarantee to individual arms. When arms are naturally grouped by certain attribute(s), we propose Bi-Level Fairness, which considers two levels of fairness. At the first level, Bi-Level Fairness guarantees a certain minimum exposure to each group. To address the unbalanced allocation of pulls to individual arms within a group, we consider meritocratic fairness at the second level, which ensures that each arm is pulled according to its merit within the group. Our work shows that we can adapt a UCB-based algorithm to achieve a Bi-Level Fairness by providing (i) anytime Group Exposure Fairness guarantees and (ii) ensuring individual-level Meritocratic Fairness within each group. We first show that one can decompose regret bounds into two components: (a) regret due to anytime group exposure fairness and (b) regret due to meritocratic fairness within each group. Our proposed algorithm BF-UCB balances these two regrets optimally to achieve the upper bound of on regret; being the stopping time. With the help of simulated experiments, we further show that BF-UCB achieves sub-linear regret; provides better group and individual exposure guarantees compared to existing algorithms; and does not result in a significant drop in reward with respect to UCB algorithm, which does not impose any fairness constraint.

Abstract

Dysarthria due to Amyotrophic Lateral Sclerosis (ALS) affects the acoustic characteristics of different speech sounds. The effects intensify with increasing severity leading to the collapse of the acoustic space of the affected individuals. With an aim to characterize such changes in the acoustic space, this paper studies the variations in band-specific and full-band spectral properties of 4 sustained vowels (/a/, /i/, /o/, /u/) and 3 sustained fricatives (/s/, /sh/, /f/) at different dysarthria severity levels. Effect of dysarthria on spectral features of these phonemes are not well explored. Statistical comparison of these features among different severities for the phonemes considered and among different vowels/fricatives for every severity level using speech data from 119 ALS and 40 healthy subjects indicate the followings. Though all band-specific and full-band features of the three fricatives and most of those features for the four vowels become statistically similar at high severity levels, certain features remain distinguishable. Spectral differences in 0-2 kHz band between /a/ and the other vowels and in the 2-6 kHz band between /a/ and /o/, /u/ persist through all severity levels. Moreover, properties of /f/ remain mostly unchanged with increasing dysarthria severity levels.

Abstract

: Gathering requirements for developing virtual reality (VR) software products is a labor-intensive process. It requires detailed elicitation of scene flow, articles in the scene, action responses, custom behaviors, and timeline of events. The slightest change in requirements will escalate the design and development costs. While most VR practitioners depend on conventional software engineering (SE) requirement-gathering techniques, there is a need for novel methods to streamline VR software development. With severe software platform fragmentation and hardware volatility, VR practitioners need assistance specifying non-volatile requirements for a minimum viable VR software product.

Abstract

Artificial intelligence is the new technological buzzword. Everything from camera apps on your mobile phone to medical diagnosis algorithms to expert systems are now claiming to be ‘AI’, and many more facets of our lives are being colonized by the application of AI/ML systems (henceforth, ‘AI’). But what does this entail to designers, users and to society at large? Most of the philosophical discourse in this context has focused on the analysis and clarification of the epistemological claims of intelligence within AI and on the moral implications of AI. Philosophical critiques of the plausibility of artificial intelligence do not have much to say about the real-world repercussions of introducing AI systems into every possible domain in the name of automation and efficiency; similarly, most of the moral misgivings about AI have to do with conceiving them as autonomous agents beyond the control of human actors. These discussions have clarified the debate surrounding AI to a great extent; however, certain crucial questions remain outside the ambit of these debates. Arguments in support of AI often take advantage of this void by emphasizing that AI systems are no different than previously existing ‘unintelligent’ technologies, thereby implying that the economic, existential, and ethical threats posed by these systems are either similar in kind as those posed by any other technology or grossly misplaced and exaggerated. In this chapter, we shall think through this assumption by investigating into the ontology of AI systems vis-à-vis ordinary (non-AI) technical artefacts to see wherein lies the distinction between the two. I shall examine how contemporary ontologies of technical artefacts (e.g., intentionalist and non-intentionalist theories of function) apply to AI. Hence, clarifying the ontology of AI is crucial to understand their normative and moral significance and the implications therefrom.

Abstract

Machine Learning models increasingly face data integrity challenges due to the use of large-scale training datasets drawn from the Internet. We study what model developers can do if they detect that some data was manipulated or incorrect. Such manipulated data can cause adverse effects including vulnerability to backdoored samples, systemic biases, and reduced accuracy on certain input domains. Realistically, all manipulated training samples cannot be identified, and only a small, representative subset of the affected data can be flagged. We formalize ``Corrective Machine Unlearning'' as the problem of mitigating the impact of data affected by unknown manipulations on a trained model, only having identified a subset of the corrupted data. We demonstrate that the problem of corrective unlearning has significantly different requirements from traditional privacy-oriented unlearning. We find most existing unlearning methods, including retraining-from-scratch without the deletion set, require most of the manipulated data to be identified for effective corrective unlearning. However, one approach, Selective Synaptic Dampening, achieves limited success, unlearning adverse effects with just a small portion of the manipulated samples in our setting, which shows encouraging signs for future progress. We hope our work spurs research towards developing better methods for corrective unlearning and offers practitioners a new strategy to handle data integrity challenges arising from web-scale training.

Abstract

Stress has a significant mental health problem of the 21st century. The number of people suffering from stress is increasing rapidly. Thus, easy-to-use, inexpensive, and accurate biomarkers are needed to detect stress during its inception. Early detection of stress-related diseases allows people to access health-care services and leads to the development of new therapies. Thus, for the early detection of stress, a biomarker would be beneficial. We aim to find if there are significant differences between stressed and non-stressed groups of participants and which brain region gets impacted during stress. We conducted experiments to acquire EEG signals to identify the most significant brain region that gets affected due to stress. This research investigates if Higuchi Fractal Dimension (HFD) extracted from EEG signals can act as potential stress biomarkers. A gamified mobile application, known as Color Word and Memory Test (CWMT), was developed, inspired by the well-known Stroop Test, to elicit mental stress at two different difficulty levels. A MUSE headband with four EEG sensors (TP9, AF7, AF8, TP10) was used to collect the EEG signals at the two difficulty levels. The Higuchi Fractal Dimension (HFD) was extracted from EEG signals acquired from 32 participants while they were exposed to stress stimuli (using the proposed CWMT application) and we then performed three experimental analyses, i.e., Analysis I, II and III. Analysis I was performed on all the 32 participants performing low and high difficulty tasks on CWMT application. Analysis II was performed on sub-groups of participants. These sub-groups were made based on score obtained by participants while using CWMT application. Analysis II aimed to find the most significantly impacted brain region during stress. For Analysis III, we used the same sub-groups as Analysis II and it aimed to identify any differences between the left and right hemispheres during stress. We conducted the statistical analysis, and p-values were calculated between the two groups (non-stressed and stressed) to detect EEG channels and brain frequency significantly associated with stress. We performed three experimental analyses (Analysis I and II are intra-hemisphere, and analysis III is inter-hemisphere). In Analysis I, we inferred that beta and alpha frequencies from the AF8 region of the brain are affected during stress. In Analysis II, we inferred that beta waves from the AF8 region are a characteristic indicator of stress. In Analysis III, we identified significant differences between the left and right parts of the brain during stress. We found a significant difference (p<0.05) between HFD value in stress and non-stressed groups in the AF8 region. Our results indicated that stressed patients have a significantly higher value of HFD in the frontal areas.

Abstract

A number of anthropogenic and natural disasters that have impacted the fragile ecosystems of the Baratang group of islands in the Andaman and Nicobar archipelago, resulting in the degradation and depletion of their natural environment and biodiversity, have been identified. It is found that, in the early 20th century, the resources of the archipelago, especially Baratang Island, were exploited for infrastructure and plantations, resulting in a decrease in forest cover and an increase in anthropogenically transformed landscapes. However, the most recent available information on land use (LU) and land cover is only from 2013. This study aims to understand the dynamics of LU and land cover change in the Baratang Island group using multitemporal satellite data for 1976, 1993, 2004, 2014, and 2023. Using visual interpretation techniques, spatial and temporal change maps were created to represent the different LU and land cover features in the study area. An analysis of these maps revealed an interesting trend: there is an increase in forest area with a simultaneous decrease in settlements and agricultural land, which contradicts the findings of previous studies. This work has identified several factors contributing to this trend, namely natural regeneration, a reduction in logging and human activities, government policies and laws, and the implementation of conservation measures by the forest department. However, the evidence on mangrove decline attributed to both human activities and natural disasters is consistent with the results of other studies. This study emphasizes the importance of regularly monitoring the island’s landscape dynamics to understand changes in LU and land cover and to take appropriate conservation measures.