Zine making with small data
思辨设计、产品设计
课程名称
CONVERSATIONS WITH MACHINES: LECTURE & WORKSHOP SERIES
合作者
Individual
项目年份
2024
设计构思
用餐体验不仅取决于食物或用餐仪式,还取决于它所唤起的个人记忆。即使是与家人、朋友或爱人共进的每一顿家常便饭,都有着深刻的意义。我们想要探究的是我们如何以有形的产品捕捉这些时刻并与他人分享。

用餐行为将我们与个人回忆联系在一起,并激发我们对食物、文化和餐桌故事的思考。受以下作品的启发,我们发现桌布可以作为记录我们用餐行为的画布, 例如我们坐在哪里、吃了什么、做了什么以及与谁分享这些时刻等等。这种媒介利于折叠、包装和赠送,使其可以成为一种珍贵的邀请。我们邀请收到它的朋友作为客人与我们一起庆祝这些亲密而珍贵的时刻。
case study
设计流程
通过头脑风暴,我们筛选出三个想法:

- 随时随地的用餐体验
- 利用投影来增强用餐体验
- 设计餐具来模拟厌食的具身体验。  

我们最终的想法是前两个想法的结合:设计并制作一种可以根据不同的餐桌尺寸和摆放方式进行变换的桌布。我们希望引入第二个想法中探讨的文化交流元素,并借鉴我们在家与家人一起用餐的方式。
ideation 01
ideation 02
我们的第一个原型是可以覆盖整个餐桌的药丸状桌布。餐布上标有盘子应该摆放的位置。我们缝上纽扣并画上折叠线,这样用户就可以根据提示将桌布折叠起来,以满足各种安排和客人数量的需要。
prototype 1 01
prototype 1 02
然而我们发现,在制作这些可折叠桌布的过程中,我们失去了该项目的初衷:与他人分享我们的私人用餐体验。盘子摆放的限制让它变得刻板无趣。我们探索过使用不同类型的织物来实现不同的摆放方式,但最终还是回到了最初的想法。我们决定放弃制作通用桌垫的想法,而是简单地制作一块专为一人的定制用餐体验。
在制作过程中,我们用不同颜色的布划分出不同家庭成员的区域,并在每个盘子上贴上他们的名字。我们还划分出手机和平板电脑的摆放位置,上面有他们吃饭时经常观看的节目或使用的应用程序的二维码。角落里的指南针并不是地理位置,而是电视机(家庭聚餐仪式的中心)相对于餐桌的位置。可爱的插图为菜肴划定了灵活的区域。
pattern design 01
此外,我们还制作了一份手册说明桌布上描绘的用餐体验,包括家庭成员介绍、餐具、器皿和配件的图例,以及日常用餐场景。我们鼓励收到这份礼物的人通过模仿用餐场景,身临其境地体验桌布上的生活。
最终的产品是一次真正意义上的合作,从缝纫、图案设计到绘画,每个团队成员都参与其中。我们一起努力保存这些记忆,就好像我们自己也是这个家庭的一员。
fabrication process 01
fabrication process 02
产品展示
最终的桌布上缝制了不同颜色的布料,代表不同的家庭成员的用餐区域。我们选择Brie的家庭作为这块桌布的模板。桌布的东南角是她弟弟的就餐区域,因为他在自己的房间里吃饭,这个角落可以通过纽扣拆卸和组装。收到礼物的人还可以扫描iPad 区域和悬挂的二维码,查看送礼者在用餐时的所见所闻,并查阅电子版手册。
final product
Coming soon :)
THE PROJECT
An AI-driven web-based plugin that provides better communication between users and generative AI tools
Discover & UNDERSTAND
Our journey starts from the question below: As a designer, have you ever used Generative AI tools in design process? And have you ever felt misunderstood by Generative AI tools in design process?

We collected opinions from designers across different disciplines and integrated them into an affinity diagram. The takeaways from this status quo guided us to our design space: human-AI communication.
surveyinsights
define question
How might we ease the process of communication between designers and generative AI tools by fostering mutual understanding?
IMAGINE
Our proposal is to leverage a web-based plugin with language process ability. Users label the results after comparing the current outcome with their mind and iterate the suggestion on the plugin. The human-in-the-loop method will provide users preferred results by achieving consensus.
workflow
RAPID PROTOTYPING
init
Feature 1: User Domain Settings (web-based plugin)

Once the plugin is installed and logged in, the interface will prompt user to select their corresponding domain in design professions, which parameters will be applied on image generation and suggestion-making processes of the plugin.
suggestion
Feature 2: User-oriented Image Classification / Generation (web-based plugin)

Once user begin to execute text-based image generation processes, the plugin will provide users with tailored user-friendly image labels; which will allow the user to  revise the labels based on one’s interpretation of content.
iteration
Feature 2: User-oriented Image Classification / Generation(web-based plugin)

Image re-generation based on revised image classification labels will help users acquire updated contents more precisely; while helping the algorithm to learn user’s domain knowledge and pattern - providing data for the algorithm to improve performance overtime.
personalization
Feature 3: User Profile Settings (dedicated web interface)

The web-based plugin also provides a standalone website that allows the user to retrospectively review domains and labels associated with their profile, recycling previously removed parameters while gaining more insights about their interaction patterns.
PRECEDENT STUDY
We wanted to build an object and experience that would embody what it would be like to be underground. We explored diorama as a possible final product as we wanted to design an experience, but also build a physical prototype. We took heavy inspiration from underground, ant colony dioramas.
precedents
PROCESS & DEVELOPMENT
Our initial sketch consisted of three levels: underground, ground, and canopy. Both the canopy and the ground level would have holes cut through the base, which would allow us to drip water through to simulate rainfall. The trees we build would play a vital part, not just in terms of aesthetics but also in terms of the experience itself. We wanted to design it so that the water that falls from the canopy would drip down, saturate the tree and its roots, carrying the water all the way down to the underground level. We also designed a staircase inviting guests to imagine what it would be like to go down into a cave, covered with roots, soil, and the countless creatures hidden underneath.
sketch
For the underground level, we built the ‘cave’ with black aluminum foil, poured soil over around the edges, and punched holes on the ceiling through which we snaked through tube cleaners. This would carry the water from above ground to below ground, once saturated. We also cut a large hole in the front panel, inviting guests to reach in and touch the soil as the water drops from the cave’s ceiling. We wrapped the black foil over and around the opening in the center, to help keep the soil contained and sealed. The cotton was threaded through thin copper wires which hung over the two side walls.
interaction
FINAL PRODUCT
The final model is completed with organic soil and manmade trees and clouds. The environment comes to life in our film through sprinkling water over the clouds and seeping into the soil through the tube cleaners. A play on both the Wildcard and Synthesis, this experience allows one to feel the senses of the Earth 6 ft under.
final 01
final 02
final 04
final 03
设计构思
我选择了我与 Amazon Alexa 的对话记录作为制作这本手册的数据。我和三个室友合租,没有在房中安装智能家居设备。因此,我与它的交流大多是播放音乐和早期的闹铃。 虽然我和它相处了近一年,但其中的数据并不多。这次工作坊给了我一个很好的机会来反思我与它的关系 - 一个陪伴我度过无数日夜的伙伴。

我在工作坊的前一天下载了数据,并将其处理成适当的格式:内容、日期和时间。我从最近的时间看起,然后发现了许多有趣并且耐人寻味的对话。我不是英语母语者,而Alexa 也不像目前最先进的大语言模型那样“聪明”。这些与这台设备建立关系中的趣闻和误解启发了我。我把所有对话的情景分为三类:

类型1: 正确识别

类型 2:Alexa识别错误:由于语音识别能力有限,Alexa 无法读取其技能集和学习算法之外的模糊命令。在这种情况下,它无法解析命令并显示:

        "Audio could not beunderstood."

但更常见的情况是类型 3:我的问题:语音录制只有在按下录制按钮时才起作用,但我的英语水平无法帮助我在这么短的时间内想出该说些什么,所以它经常什么也录不下来,显示为

        "No text stored."

有时,我也会高估它的能力,所以当我说了一些它显然无法识别的话时,它将其解析为:

        "Audio was not intendedfor this device."

通过观察上述类别的数据,我开始会对这些对话的发生时间、频率随时间的变化以及对话发生后我对和Alexa 的态度和它对我的印象有何变化感到好奇。

因此,我决定在手册的信息可视化中重点关注以下两点:
1. 以上三类每次对话的发生时间轴
2. 以及这些类别变化的趋势图

但过早地进行定量分析会抑制我的创造力,所以我试图通过浏览数据集和标注数据来想象这些趋势。在这个过程中,我发现自己就像一个侦探,对话的内容和时间唤醒了我当时的记忆。这些生动的回忆触动了我,因此我决定通过与手册的互动重现这种体验。
data before
data after
设计流程
有了明确的目标,我开始制作手册的原型。呈现时间线的最佳方式是一张卷轴,而手册的可阅读性限制了它的折叠形式。我设计了一组记忆卡片作为解密钥匙,引导人们找到卡片上故事发生的时间。手册的封面不仅起到保护作用,也可以作为阅读这两个可视化图表的良好媒介。

确定了交互的实体形式后,我开始用 Python 制作信息可视化图表并进行了图例与整体布局设计。
proto 02 - 1
proto 02 -2
第一个功能原型
python
ai
数据可视化和处理
proto 03 - 1
proto 03 -2
第二个原型:迭代测试
最终产品
最终的成品是手册的高保真原型,这里的颜色代表了我与Alexa两个叙事主体,也暗示了一天的时间(白天/黑夜)。封面被设计成一把尺子,并标记了两者重要的态度转换时刻。读者只需自己在地图上找到它们即可。
final overview
final 01
final 02
takeout
map
me
front
back
alexa
设计反思
由于时间有限,设计没有进行足够的迭代,我也没有从观众那里得到足够的反馈。因为图表的篇幅限制,它无法提供额外信息,因此对观众来说从图表中找到的完整故事可能难以理解。另外,趋势图与我的预期完全不同,看来它不太了解我,我也没有更熟练地掌握如何操作它。最后,Alexa的数据集没有存储自己的回答,数据中只有我提出的问题。尽管我从 Alexa 那里记住了一些有趣的答案,但由于缺乏数据,我无法表达出来。 总之, 这次工作坊为我提供了一个让我反思自己与个人物品之间的关系,并以平等的心态思考它们的机会。