香京julia种子在线播放

Light is the dominant data source in the Universe; therefore, our sense of sight pervades historical astronomy. For millennia, humans explored the sky with the unaided eye until the invention of the telescope (1608) provided a deeper view of the cosmos. In 1851, the first daguerreotype of a Solar eclipse captured the Sun's light, the first astronomical image (Figure 1). In the twentieth century, the quality and quantity of astronomical data experienced massive growth; during World War II, the development of “false color” enhanced astronomical image interpretability (Mapasyst, 2019), and digital data capture with charge-coupled devices (CCDs; Rector et al., 2015) created the potential to collect tremendous amounts of data.

Newly launched space-based observatories captured many of these large data sets, which revolutionizing astronomy's cultural impact (Taterewicz, 1998) by recording the energy, time, and location of photons emitted from cosmic objects. Space telescope data, delivered through NASA's Deep Space Network,¹ is vast and beyond typical human comprehension, from all-sky stellar mapping to imaging gargantuan galaxies (Marr, 2015; Arcand and Watzke, 2017). Images are central to understanding the scope and significance of such vast catalogs (Smith et al., 2011; Arcand et al., 2019), creating an information landscape where archival data have immense value and visualized data hold additional interpretive utility (Hurt et al., 2019).

In the late twentieth century, CCDs began to specialize, observing and transposing light outside visible wavelengths, extending the observable Universe once more. Here, data visualization became essential. NASA's “Great Observatories” [originally the Hubble Space Telescope, Compton Gamma Ray Observatory, Chandra X-ray Observatory, and Spitzer Space Telescope (National Aeronautics and Space Administration, 2009), and now the James Webb Space Telescope] rely on the conversion of their digital transmissions into images (Rector et al., 2015). These data can be combined into aesthetically beautiful images and shared with non-experts (English, 2016), relaying important scientific messages. With the internet, astronomers may couple these innovatively visualized data with worldwide public dissemination (Rector et al., 2015).

Image aesthetics is essential in aiding data comprehension (Arcand et al., 2013; Rector et al., 2015), and the nature of the visualization varies depending on the audience; for example, a plot communicating a scientific result in a journal or a scientifically accurate but aesthetically pleasing image to communicate publicly (DePasquale et al., 2015). Astronomers must alter images for light outside the visual spectrum regardless of the audience; therefore, it is feasible to use other methods of data vivification (Sturdevant et al., 2022) to elicit new understanding or meaning-making.

Studies on astronomical image processing (Rector et al., 2007, 2017; DePasquale et al., 2015), most notably the use of color (Smith et al., 2011, 2015, 2017; Arcand et al., 2013), have shown image creators must explain the translation process for non-expert audiences and that lay explanations perpetuate the public's confidence in the image's scientific nature and authenticity, even when they depict components concealed from human eyes, indicating the same is possible when translating visual data to sound.

Sonification is often defined as mapping data to sound to represent information using non-speech audio, the sonic counterpart to visualization (Kramer et al., 2010; Sawe et al., 2020). For decades, image sonification has communicated spatial information to blind or low-vision (BLV) individuals (Meijer, 1992; Yeo and Berger, 2005; Zhang et al., 2014). It is also used for audio alerts (i.e., Geiger counters; Kramer et al., 2010), research (i.e., studying brain wave changes; Parvizi et al., 2018), communications and education (i.e., multimedia cosmology sonification; Ballora and Smoot, 2013), increasing accessibility (i.e., the sonification of visual graphs; Ali et al., 2020), and for art and entertainment (i.e., compositional works based upon neurological or cosmic particle data; Sinclair, 2012).

In recent years, sonification has become more present in the world of astronomy as communicators and researchers alike try to understand how best to engage the BLV community and allow astronomers to conduct their research using sound alone; see Harrison et al. (2022) and Noel-Storr and Willebrands (2022) for discussions on this topic. Generally, a primary goal of astronomy sonification relates to public engagement (Zanella et al., 2022); however, research investigating active listening by astronomers for data exploration and analysis exists (Alexander et al., 2010; Diaz-Merced, 2013). A notable project in this area is the Audible Universe Project by Misdariis et al. (2022), which aims to create a dialogue between sonification and astronomy. Sonified celestial objects include pulsars, the cosmic microwave background, and solar eclipses (McGee et al., 2011; Ballora, 2014; Eclipse Soundscapes, 2022). In the past decade, astronomy-related sonifications have increased dramatically (Zanella et al., 2022), with more than 98 projects globally representing a third of the Sonification Archive (Lenzi et al., 2021).

The most common approaches to sonification are audification and parameter mapping. With audification, data are translated and mapped directly to audio, so the relevant frequencies fall within hearing range; for example, the sonification of gravitational wave signals recorded by the Laser Interferometer Gravitational-Wave Observatory (LIGO) collaboration (Abbot, 2016); which interprets space-time fluctuations caused by passing gravitational waves as an audio signal, allowing us to hear black hole mergers in real-time (Gravitational Wave Open Science Center, 2022).

For parameter mapping, aspects of the data control specific audio parameters. For example, a star's brightness fluctuations or hues of color may control the frequency so that unique features can be identified (Astronify, 2022); it is common to map brightness to pitch as our ears are more sensitive to variations in frequency than volume. The inverse spectrogram is the most common image mapping technique (Sanz et al., 2014), mapping one axis to time and the other to frequency, allocating a corresponding brightness for each pixel to control volume. For example, we may scan an image from left to right, with different pitches indicating the vertical position of objects. Due to the wide variety of possible mappings, an explanation of the process is vital for the listener to extract meaningful information.

Sonification provides astronomy communicators a new avenue to engage the public, particularly BLV communities, traditionally excluded from engagement. Furthermore, sonification is advantageous as a research tool (with or without visually presented information) because listening to data exploits the auditory system's exquisite sensitivity to pattern variation over time, whether perceived as discrete rhythms or changing pitch (Walker and Nees, 2011). In addition, because sounds are multidimensional, we may encode many parallel data streams by mapping each to a different audio dimension (pitch, volume, timbre) or control multiple simultaneous audio streams so our ears can either listen holistically or focus on one stream at a time (Fitch and Kramer, 1994). Finally, we can render each layer of a multi-wavelength image as a separate audio stream (a different note or instrument) to explore the relationship between wavelengths of data. As such, sonification has excellent potential for stimulating curiosity, increasing engagement, and creating an emotional connection with data.

In recent years, NASA has released several sonification projects, showcasing several decades of data (National Aeronautics and Space Administration, 2020, 2022; and others). In 2020, NASA's Chandra X-ray Observatory launched 'The Universe of Sound,² providing bespoke audio representations of astronomical datasets for non-expert audiences and working with BLV representatives to create and test sonifications. In this work, we analyze survey data to investigate the effectiveness of sonifications, particularly for BLV communities. This paper represents the first study to explore responses to astronomical sonifications from the BLV community and compare these responses to the experiences of sighted participants.

Before we discuss the results of our work, it is essential to acknowledge our positionality within the context of this study. For some authors, our motivation to explore this topic is shaped by a personal connection to the disability community through lived experience (either in the BLV community or the broader disability community); for others, the motivation lies in the desire to explore alternate data vivification processes and understand how to communicate science to the public effectively. Our own experiences have led us to believe sonification is a positive tool for education and research, and we remain mindful of this bias throughout our analysis. Finally, although our team represents a range of perspectives within the disability and astronomy community, we acknowledge that we remain limited by our lived experiences as a group of majority white individuals living in North America and Canada.

The research participants were a convenience sample of respondents (18 years and older) to an online survey. We solicited participants from websites including Chandra³ and Astronomy Picture of the Day (APOD),⁴ digital newsletters, social media sites such as Facebook and Twitter for Chandra⁵ and APOD,⁶ and the social media and contacts of the principal investigator (PI). Further distribution occurred through additional social media sharing. The survey was active on SurveyMonkey⁷ for 4 weeks beginning February 24, 2021. We note that SurveyMonkey surveys are compatible with assistive software typically used by the BLV community, particularly screen readers and screen magnification.

The Smithsonian Institutional Review Board⁸ determined that this survey was exempt research under Smithsonian Directive 606.⁹ The survey started with a participant consent form in which choosing to continue with the survey equaled consent. We provided no compensation to survey participants or dissemination partners.

We chose three sonifications from the six available at the time of the study at NASA's Universe of Sound website, choosing those that best represent the collection available for their variation of instrumental vs. synthetic sounds and how they track the visual data—left to right, top to bottom, or radially. Survey participants were presented with the sonifications and their accompanying videos to experience as they were able, followed by short text descriptions (screen-reader adaptable) for each of the represented astronomical objects (the Galactic Center, Cassiopeia A, and the Chandra Deep Field South). The sonifications played in the same order, without counterbalancing, starting with the sonification that used non-synthesized sounds, followed by two more complex sonifications. We provide the details of each sonification at the companion GitHub¹⁰ and highlight the key points, along with links to the sonifications, in Table 1.

Our survey began with five demographic questions¹¹ (age, gender, education level, self-rated knowledge of astronomy, and whether the participant identified as BLV).

Participants were then asked to experience the sonifications and after each, respond to a set of statements using a Likert scale. Each statement began with:

The five statements were:

The scale provided the following options: 1 (Disagree Strongly), 2 (Disagree), 3 (Neutral), 4 (Agree), 5 (Agree Strongly).

Following this section, the participants were asked about their overall experience, first:

Then, they were asked to rate the following three statements, using the same Likert scale from the first section:

Finally, they were asked two open-ended questions (which allowed for full sentences). These were:

Once the survey closed, we exported the data and cleaned and analyzed the 4,346 responses using Python. We removed the entry of one participant who took from March until July 2021 to complete the survey and all responses in which participants did not indicate whether they were BLV or sighted or answered fewer than three non-demographic questions. This cleaning ensured we could compare the results of the BLV and sighted groups for those who engaged with the sonification questions. We removed identical entries by comparing Internet Protocol (IP) addresses and demographic questions. For repeat entries, we kept only the most recent response. Cleaning yielded 3,184 participant responses. See the Appendix for the demographic breakdown of the cleaned sample.

Figure 2 displays self-reported knowledge of astronomy, divided by the BLV (blue) and sighted (red) participants. The apparent contrast in size of the two demographic groups is discussed in Results (Section 3) and Future Work (Section 6.1).

Regarding additional demographics, we note a slight majority of male-identifying participants (57.1%). Participant ages spanned from 18 to 24 years (21.6%) to 65 years and older (16.3%); there is a slight predominance of younger participants, but all age groups are represented at above 10% of the total. Likewise, the self-reported education level of participants ranges from those who completed some of high school to those with advanced postgraduate degrees (i.e., doctorate, LLB, or MD); however, those who completed some of high school were the least represented group (3.5%), with most participants (61.3%) having completed an undergraduate degree or higher. We refer the reader to the tables in the Appendix for a complete breakdown of participants' demographic data.

Figures 3–6 display responses to the sonification prompts, separated into BLV and sighted participants, and displayed in order of sonification from left to right (Galactic Center, Cassiopeia A, and the Deep Field). We performed 2-sided Kolmogorov–Smirnov (K-S) tests on each set of distributions using the Python module Scipy's ks_2samp¹² function. We elected to use the K-S Test, a non-parametric test, because we did not expect the distribution of responses to our survey to be normal, which visual inspection of the data confirmed. We define a p-value of < 0.05 as evidence against the null hypothesis and a p-value < 0.001 as strong evidence against the null hypothesis. Although our sample sizes differ between the BLV and sighted groups, the 2-sided K-S test can accommodate these differences while maintaining validity, and the default “auto” parameter used can handle small sample sizes. However, this sample has more significant uncertainty due to the smaller number of BLV participants. The results of the K-S tests and p-values for all responses can be found in Table 2.

Figure 3 shows participant ratings for the prompt: “I enjoyed this experience.” Generally, all participants reported enjoying the sonifications, with the majority selecting 4 (Agree) and 5 (Agree Strongly). A higher number of BLV participants selected 5 for the Galactic Center and Cassiopeia A, whereas the enjoyment ratings for the Deep Field are almost identical for both groups. Cassiopeia A shows the most extensive range of ratings, and although more BLV participants selected the highest rating, the p-value does not suggest a statistically significant (0.620) difference between the two groups; however, the p-value is significantly lower than for the other sonifications (0.997 and 1.000, respectively).

Ratings for the prompt: “I learned more about the [title of sonification, i.e., Galactic Center] through this experience,” are shown in Figure 4. In general, most participants felt they learned something about the cosmic sources. For each sonification, the BLV group rated learning more, particularly for Cassiopeia A, where the low p-value (p = 0.02407) suggests a statistically significant difference in the responses of the BLV and sighted groups. On average, both groups claimed to learn most about the Deep Field.

Figure 5 shows ratings for the prompt: “Hearing the sounds enhanced my experience.” Generally, participants felt that adding audio to the astronomical images enhanced their experience, particularly for the Deep Field. Interestingly, all the p-values are high, suggesting that both the sighted and BLV participants found their experience enhanced to the same extent.

The responses to the prompt “Watching the video enhanced my experience” are depicted in Figure 6. The K-S test results for all three objects indicate that the responses from both groups are statistically similar, with the p-values for Cassiopeia A and the Deep Field suggesting the highest similarity. We note that across all three objects, both groups infrequently responded that they disagreed (1 or 2), with 4 and 5 being the most common response, suggesting both groups generally found the video to be a beneficial addition to their experience.

The prompt: “I trust that this representation is faithful to the science data,” is displayed in Figure 7. The frequency of 4 and 5 ratings indicating agreement suggests that participants believed the sonifications were scientifically accurate. The high p-values (all > 0.9) suggest no evidence that the trust levels differed between the groups.

Figure 8 shows the ratings for three prompts given after listening to the sonifications: “I am motivated to listen to more [sonifications],” “I am interested in learning more about our Universe,” and “I want to learn more about how others access information about the Universe.” The p-value for the left-hand histogram implies no difference between the groups regarding whether they wanted to listen to more.

The distribution of ratings regarding interest in continued learning about the Universe and how others access this information differed significantly between the groups (p-values < 0.05). In both cases, BLV participants responded 5 (Agree Strongly) more frequently than sighted users. Although sighted participants responded 5 with a lower frequency to these two prompts, the most common responses were still in agreement (4 and 5), indicating that sighted participants were also interested in learning more. Of the 2,203 sighted respondents to these questions, 1,708 responded in agreement to the prompt regarding being motivated to listen more, 1,798 responded in agreement to the prompt about being interested in learning more about the Universe, and 1,710 responded in agreement to the prompt regarding wanting to learn more about how others access information about the Universe. One sighted participant who answered the other two final prompts did not respond to the prompt about wanting to learn more about the Universe.

Figure 9 shows a word cloud,¹³ displaying the terms participants used to respond to, “List up to three words to describe your emotional response to these data sonifications.” Word size corresponds to their frequency of use. For terms that pertain to a positive experience, the number of instances is as follows: The combined terms “curious” and “curiosity” totaled 329 instances, and “calm” showed the highest number of entries for a single term (326), followed by “interesting” (243). Additional terms included “relaxed/relaxing” (216), “amazed” (169), “wonder” (168), “beautiful” (133), “peaceful” (132), and “awe” (124). Negative terms also appeared, including (but not limited to): “Boring/bored,” “confused/confusion,” “stress,” “disturbed,” “pointless,” “gimmicky,” and “scary,” but these appeared with far less regularity.

The first open-ended question asked, “What recommendations do you have to help the scientific community create better listening experiences?” There were no character limits imposed on the answers. Using manual inductive coding (Chandra and Shang, 2019) through sampling and re-coding, we collated responses into seven broad categories: general comments, technical, scientific, musical, educational, sensory, and accessibility-related, chosen based on the themes seen in the responses. There were 1,417 responses (removing all non-descriptive responses such as blanks or symbols [i.e., ///]). A complete summary of the responses is available in the data repository.

Amongst the responses, we noted two frequent themes; the first, a common misunderstanding of sonification, both at its conceptual level (e.g., “Include actual sound from space”) and in the context of interpretation (e.g., “I don't fully understand the relationship between the sounds and what we are seeing”). These comments suggest an unfamiliarity with sonification as a form of data representation, and the audience may require more background to interpret this representation correctly. We teach students to read graphs and charts visually, so education in sonification might likewise be necessary, echoing the suggestions of Fleming (2023).

The second theme noted is the frequent suggestions regarding the assignment of pitches and other audio parameters to the data, ranging from musical suggestions (e.g., “Please don't stick to the equal temperament system in sound reproduction, so much scientific information is lost or misrepresented that way. Also, why link different things to different pitches suggesting differences in quality better represented by different timbre?”) to responses tagged as scientific (e.g., “If you're going to assign a sonification to individual elements, I think you're going to have to find a better way to differentiate between them than to just change the note on the scale. Maybe brainstorm a way to differentiate between them based on atomic weight or outer electron shell, assigning a sonification to the sounds the orbits might make.”). These responses reflect the question of standardization in sonification, much the same as the standards for visual data representation: how we represent different images or data types in a way that is both interpretable across different sonifications and auditorily pleasant. These standards could improve the feasibility of sonification education.

The second open-ended question asked, “If the person who created these data sonifications were here, what question would you ask them?” We coded these responses into the same seven categories. There were 1,656 responses after removing non-descriptive responses. Across the categories, many questions involved the purpose of these sonifications (e.g., “is the goal enchanting soundscapes or information transfer or enhancing information acquisition in non-visual.” and “By a glimpse to photo we can all have these information at once. So what is the use of this?”). Other questions inquired how the audio parameters were mapped to the image data (e.g., “Did you select the frequency distributions to try and make the sonifications tuneful, or are they evenly (linearly, logarithmically) spaced across the audio spectrum?”). The ubiquity of these questions indicates the relative novelty of and lack of familiarity with sonification as a data representation tool among general audiences.

This study represents a valuable contribution to accessibility in astronomy; however, it is not as rigorous as desired. We selected participants via a convenience sample, where they voluntarily chose to complete our survey after receiving the link from a newsletter (Chandra or APOD) or astronomy-related social media account. Due to the voluntary nature of participation, those involved may be more interested in astronomy and have a base of knowledge, possibly affecting their interpretation of the sonifications. By formulating a questionnaire that (in part) attempts to obtain opinions on sonification products produced by the authors, we may have introduced a social desirability bias, potentially causing participants to respond more favorably to the sonifications. A complete analysis of this effect is outside this work's scope, but we may consider it more thoroughly in future publications.

Our most significant limitation was the lack of BLV respondents, with the smaller sample size resulting in increased uncertainty in the distribution of their responses. Finally, our survey includes a United States-heavy participant distribution due to how we circulated the survey.

Future work must focus on the active engagement of BLV participants while recognizing and accommodating the wide range of visual impairments within this non-homogenous group. Efforts could employ different sampling techniques to recruit a larger sample, particularly for a range of BLV individuals with a scope of astronomy familiarity. BLV participants without astronomy familiarity provide insight into how intuitive sonifications are, whereas participants with more familiarity can share how well sonifications match or enhance their understanding of the objects.

We acknowledge that the BLV category spans a broad range of sight loss that this study does not explore or quantify. Future research should ask participants to comment on the usefulness of the images accompanying the sonification as a proxy for measuring their functional vision. Researchers could also collect data on the accessibility software used while completing the survey (e.g., screen magnification, screen readers, Braille displays, and other methods) to understand whether BLV participants access the survey visually or often visually access their computers. Furthermore, one could ask for feedback regarding the visualizations to improve the accessibility of these data representations to those with low vision.

Astronomy communicators must continue to address and resolve misunderstandings of the sonification process by improving accompanying descriptions of the techniques used. These updates must consider the lens of trust in science and be mindful of creating minimal opportunities for miscommunication. To understand this better, we must capture data on the number of times a participant plays a sonification, providing a more objective measure of comprehensibility, intuitiveness, enjoyment, and a desire to understand.

Further studies could gauge the self-reported knowledge of music and technology. Many participants gave feedback on the musical quality, indicating an understanding of music theory, and many also gave technical feedback (bearing in mind that some technical proficiency is required to access the survey).

Although we collected participants' ages, we did so primarily to compare the representativeness of our sample to the overall U.S. population (see Table 3 in the Appendix) and provide thorough information regarding our participants. Future work could explore whether age correlates with enjoyment, self-reported learning, trust, and overall responses to the sonifications, although analysis of this is beyond the scope of our work. We could also explore the role of misconceptions with age. Future studies should be mindful that some participants may have hearing loss, which we do not report here, and could impact the response to sonifications. Hearing loss is more likely with increased age and could further impact the relationship between age and response to sonifications. Other demographic questions, in particular self-reported knowledge of astronomy, could also reveal interesting relationships with responses to sonification and can be explored in the future.

Finally, this work could extend to investigate actual learning outcomes, as opposed to self-reported learning (as in this study). However, this is outside this paper's scope and would involve a participant and control group learning with and without access to sonification.

Input from the broader community is invaluable, and we are encouraged by the recommendations received and excited to implement them into new work. We look forward to collaborating with others throughout astronomy and related fields to make as much data available to as many people as possible. Additional resources are available for this paper on a companion GitHub (see text footnote ¹⁰) and a frozen Zenodo¹⁶ repository.